United States
Environmental Protection
Agency
Water Distribution System Analysis
Field Studies, Modeling
and Management
A Reference Guide for Utilities
Field Studies
ACM12
Modeling
Flow gage
O Conductivity meter(CM01-20|
m Injection point
CMOS CM04
Management
D Near Entry Point
if/ D Average Residence Time
D HighTotalTrihalomethanes
D High Haloacetic Acids
-------
EPA/600/R-06/028
December 2005
Water Distribution System Analysis:
Field Studies, Modeling and Management
A Reference Guide for Utilities
U. S. Environmental Protection Agency
Office of Research and Development
National Risk Management Research Laboratory
Water Supply and Water Resources Division
Cincinnati, Ohio
Recycled/Recyclable
Printed with vegetable-based ink on
paper that contains a minimum of
50% post-consumer fiber content
processed chlorine free
-------
A Reference Guide for Utilities
Notice
Any opinions expressed in this document/reference guide for utilities are those of the author(s) and
do not, necessarily, reflect the official positions and policies of the U.S. Environmental Protection
Agency (EPA). Any mention of products or trade names does not constitute recommendation for use
by EPA. This document has been reviewed in accordance with EPA's peer and administrative review
policies and approved for publication.
-------
A Reference Guide for Utilities
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the Nation's
land, air, and water resources. Under a mandate of national environmental laws, the Agency strives
to formulate and implement actions leading to a compatible balance between human activities and
the ability of natural systems to support and nurture life. To meet this mandate, EPA's research program
is providing data and technical support for solving environmental problems today and building a
science knowledge base necessary to manage our ecological resources wisely, understand how
pollutants affect our health, and prevent or reduce environmental risks in the future.
The National Risk Management Research Laboratory (NRMRL) is the Agency's center for investigation
of technological and management approaches for preventing and reducing risks from pollution that
threaten human health and the environment. The focus of the Laboratory's research program is on
methods and their cost-effectiveness for prevention and control of pollution to air, land, water, and
subsurface resources; protection of water quality in public water systems; remediation of contaminated
sites, sediments and groundwater; prevention and control of indoor air pollution; and restoration of
ecosystems. NRMRL collaborates with both public and private sector partners to foster technologies
that reduce the cost of compliance and to anticipate emerging problems. NRMRL's research provides
solutions to environmental problems by: developing and promoting technologies that protect and
improve the environment; advancing scientific and engineering information to support regulatory
and policy decisions; and providing the technical support and information transfer to ensure
implementation of environmental regulations and strategies at the national, state, and community
levels.
This publication has been produced as part of the Laboratory's strategic long-term research plan. It is
published and made available by EPA's Office of Research and Development to assist the user
community and to link researchers with their clients.
Sally Gutierrez, Director
National Risk Management Research Laboratory
-------
A Guide for Utilities
Table of Contents Overview
Overview of distribution system
infrastructure, general water quality issues
and management approaches
Chapter 1
Summary discussion of various hydraulic
and water quality modeling tools
Chapter 2
Outline for planning and execution of a
tracer study to perform distribution system
evaluation
Chapter 3
Techniques for calibration and validation of
distribution system models
Chapter 4
Framework of available options for
monitoring distribution system water quality
Chapter 5
Introduction to the use of geospatial
technology for water distribution systems
Chapter 6
Real world applications of distribution
system modeling approaches
Chapter 7
IV
-------
A Reference Guide for Utilities
Table of Contents
Notice ii
Foreword iii
Table of Contents Overview iv
Table of Contents v
List of Tables ix
List of Figures ix
Acronyms and Abbreviations xii
Acknowledgments xv
1.0 Introduction 1-1
1.1 Distribution System - Infrastructure Design and Operation 1-3
1.1.1 Key Infrastructure Components 1-3
1.1.1.1 Storage Tanks/Reservoirs 1-3
1.1.1.2 Pipe Network 1-3
1.1.1.3 Valves 1-3
1.1.1.4 Pumps 1-4
1.1.1.5 Hydrants and Other Appurtenances 1-4
1.1.2 Basic Design and Operation Philosophy 1-4
1.1.2.1 Pipe-Network Configurations 1-5
1.1.2.2 Multiple Source Configuration 1-5
1.1.2.3 Impact of System Design and Operation on Water Quality 1-5
1.2 Water Quality Problems and Issues 1-6
1.3 Regulatory Framework 1-7
1.4 Assessmentand Management of Water Quality 1-8
1.5 Advanced Tools for Water Quality Management 1-11
1.6 Report Organization 1-11
1.7 Summary 1-11
References 1-13
2.0 Modeling Water Quality in Drinking Water Distribution Systems 2-1
2.1 Distribution System Network Hydraulic Modeling 2-1
2.1.1 History of Hydraulic Modeling 2-1
2.1.2 Overview of Theoretical Concepts 2-2
2.1.3 Basic Hydraulic Model Input Characterization 2-3
2.1.3.1 Pipe Network Inputs 2-3
2.1.3.2 Water Demand Inputs 2-3
2.1.3.3 Topographical Inputs 2-5
2.1.3.4 Model Control Inputs 2-5
2.1.3.5 Extended Period Simulation (EPS) Solution Parameters 2-5
2.1.4 General Criteria for Model Selection and Application 2-6
2.1.4.1 Developing a Basic Network Model 2-6
2.1.4.2 Model Calibration and Validation 2-6
2.1.4.3 Establishing Objectives and Model Application 2-6
2.1.4.4 Analysis and Display of Results 2-7
2.2 Modeling Water Quality In Distribution System Networks 2-7
2.2.1 History of Water Quality Modeling 2-7
2.2.2 Theoretical Concepts for Water Quality Modeling 2-8
2.2.3 Water Quality Model Inputs and Application 2-10
2.3 Hydraulic and Water Quality Modeling Software 2-11
2.3.1 E PAN ET Software 2-11
2.3.2 Commercial Hydraulic-Water Quality Modeling Software 2-12
2.4 Additional Modeling Tools 2-12
2.4.1 Storage Modeling Tools 2-12
2.4.1.1 CFD Models 2-13
2.4.1.2 Compartment Models 2-14
2.4.1.3 Physical Scale Models 2-14
2.4.2 Transient Analysis Software 2-15
-------
A Reference Guide for Utilities
2.4.3 Optimization Tools 2-15
2.4.3.1 Optimizing Calibration 2-15
2.4.3.2 Design Optimization 2-16
2.4.3.3 Optimization of Operation 2-17
2.4.4 Probabilistic Models 2-17
2.5 Summary and Conclusions 2-17
References 2-19
3.0 Tracer Studies for Distribution System Evaluation 3-1
3.1 Planning and Designing a Distribution System Tracer Study 3-2
3.1.1 Establishing Study Objectives and Time-Line 3-2
3.1.2 Forming a Study Team 3-2
3.1.3 Defining Study Area Characteristics 3-2
3.1.4 Selecting Tracer Material 3-3
3.1.4.1 Fluoride 3-3
3.1.4.2 Calcium Chloride 3-4
3.1.4.3 Sodium Chloride 3-5
3.1.4.4 Other Chemicals That May be Added as Tracers 3-5
3.1.4.5 Naturally or Normally Occurring Tracers 3-5
3.1.4.6 Comparison of Tracers 3-6
3.1.5 Selecting Field Equipmentand Procedures 3-8
3.1.5.1 Injection Pump(s) 3-8
3.1.5.2 Tracer Storage and Dosage Rate Measurement 3-8
3.1.5.3 Distribution System Flow Rate Measurement 3-9
3.1.5.4 Field Measurement of Tracer Concentration 3-9
3.1.6 Developing a Detailed Study Design 3-10
3.1.7 Addressing Agency and Public Notification 3-11
3.2 Executing a Tracer Study 3-11
3.2.1 Procurement, Setup, Testing and Disinfection of Study Equipment 3-11
3.2.2 Installation of Field Equipment and Testing 3-12
3.2.3 Tracer Dosage and Injection Duration Calculations 3-13
3.2.4 Dry Runs and Planned Tracer Injection Event(s) 3-13
3.2.5 Real Time Field Assessments, Sampling, and Analysis 3-13
3.2.6 Equipment De-Mobilization, Initiation of Data Collection, Reduction, and Verification
Process 3-14
3.3 Tracer Study Costs 3-14
3.4 Summary, Conclusions and Recommendations 3-15
References 3-17
4.0 Calibration of Distribution System Models 4-1
4.1 Hydraulic and Water Quality Model Calibration 4-2
4.1.1 Hydraulic Model Calibration 4-2
4.1.2 Water Quality Model Calibration 4-2
4.2 Static Calibration and Dynamic Calibration 4-4
4.2.1 Steady-State Calibration Methods 4-4
4.2.1.1 C-Factor Tests 4-4
4.2.1.2 Fire-Flow Tests 4-5
4.2.1.3 Chlorine Decay Tests 4-5
4.2.2 Dynamic Calibration Methods 4-6
4.3 Manual Calibration and Automated Calibration 4-7
4.4 Case Studies 4-7
4.4.1 Case 1 - Small-Suburban, Dead-End System 4-8
4.4.2 Case 2 - Large-Suburban Pressure Zone 4-10
4.5 Future of Model Calibration 4-13
4.5.1 Calibration Standards 4-13
4.5.2 Technological Advances 4-14
4.6 Summary and Conclusions 4-14
References 4-15
vi
-------
A Reference Guide for Utilities
5.0 Monitoring Distribution System Water Quality 5-1
5.1 Establishing Monitoring Objective(s) 5-1
5.1.1 Regulatory Driven Monitoring 5-2
5.1.2 Security Related Monitoring 5-2
5.1.3 Process Control-Related Monitoring 5-3
5.1.4 Water Quality Characterization 5-3
5.1.5 Multi-Purpose Use of Monitoring Data 5-3
5.2 Monitoring Techniques 5-3
5.2.1 Manual Grab Sampling 5-3
5.2.2 Automated/Online Monitoring 5-4
5.3 Monitoring Equipment Overview 5-4
5.3.1 Physical Monitors 5-4
5.3.2 Chemical Monitors 5-4
5.3.3 Biological Monitors 5-4
5.4 Establishing Monitoring Requirements 5-5
5.4.1 Monitoring Parameters 5-5
5.4.2 Number and Location of Monitors 5-5
5.4.2.1 Number of Monitors 5-6
5.4.2.2 Optimal Monitor Locations 5-6
5.4.3 Monitor Characteristics 5-7
5.4.4 Amenability to Remote Monitoring and SCADA Integration 5-8
5.4.4.1 Online Sampling/Control Devices 5-9
5.4.4.2 SCADA or Remote Monitoring Network 5-9
5.4.4.3 Field Wiring and Communication Media 5-9
5.5 Engineering and Evaluating a Remote Monitoring System 5-10
5.5.1 Remote Monitoring System Evaluation 5-11
5.6 Monitoring Case Studies 5-11
5.6.1 Rural Community Application 5-11
5.6.2 Washington D.C. Remote Monitoring Network 5-11
5.6.3 Tucson Water Monitoring Network 5-12
5.7 Summary and Conclusions 5-14
References 5-15
6.0 Geospatial Technology for Water Distribution Systems 6-1
6.1 History of Geospatial Data Management 6-2
6.1.1 Mapping, Surveying, and Remote Sensing 6-2
6.1.2 CADD 6-4
6.1.3 CIS 6-4
6.1.4 OEMs 6-5
6.1.5 Database Management Systems 6-6
6.1.6 Facility Management 6-6
6.2 CIS Principles 6-6
6.2.1 CIS Features 6-6
6.2.2 Topology 6-7
6.2.3 Map Projections, Datum, and Coordinate Systems 6-8
6.2.4 CIS Database Design 6-8
6.2.5 Management of CIS 6-8
6.3 Geospatial Data Management in the Water Industry 6-9
6.3.1 CADD 6-9
6.3.2 CIS 6-9
6.3.3 CIS 6-9
6.3.4 SCADA 6-9
6.3.5 LIMS 6-9
6.3.6 Support Technology 6-9
6.4 Integration of Geospatial Data Management and Modeling 6-10
6.4.1 Model Integration Taxonomy 6-10
6.4.2 Issues in Integrating CIS and Water Distribution System Models 6-11
vii
-------
A Reference Guide for Utilities
6.5 Use of CIS in Water Utilities-Case Studies 6-11
6.5.1 UseofGISatLVVWD 6-11
6.5.1.1 Pressure Complaint Resolution 6-13
6.5.1.2 Water Main BreakAnalysis 6-13
6.5.2 Geo-coding for Demand Forecasting and Allocation at Denver Water 6-13
6.6 Summary 6-17
References 6-18
7.0 Real-World Applications-Planning, Analysis, and Modeling Case Studies 7-1
7.1 Analysis of Waterborne Outbreak-Gideon, Missouri 7-1
7.1.1 Gideon Case Study Overview 7-1
7.1.2 The Gideon Water System Setup 7-1
7.1.3 EPA Field Study 7-2
7.1.4 Distribution System Evaluation 7-2
7.1.5 Case Study Summary and Conclusions 7-3
7.2 Reconstructing Historical Contamination Events - Dover Township (Toms River), NJ 7-4
7.2.1 Case Study Overview 7-4
7.2.2 Overall Modeling Approach 7-5
7.2.3 Simulation Techniques 7-6
7.2.4 Simulation Results and Conclusions 7-7
7.3 Application of Water Distribution System Modeling in Support of a Regulatory Requirement 7-9
7.3.1 IDSE Requirements Overview 7-10
7.3.2 Example Application of Modeling in the IDSE Process 7-10
7.4 Use of Water Distribution System Models in the Placement of Monitors to Detect Intentional
Contamination 7-13
7.4.1 Red Team-Blue Team Exercise 7-13
7.4.2 Application of Optimization Model 7-14
7.4.3 Case Summary 7-15
7.5 Case Study- Use of PipelineNet Model 7-15
7.5.1 Overview 7-16
7.5.2 Model Calibration 7-16
7.5.3 Monitoring Site Location Methodology 7-16
7.5.4 Response and Mitigation Tools 7-17
7.5.4.1 Consequence Assessment Tool 7-17
7.5.4.2 Isolation Tool 7-17
7.5.4.3 Spatial Database Display Tool 7-17
7.5.5 Case Summary 7-17
7.6 Use of Threat Ensemble Vulnerability Assessment (TEVA) Program for Drinking Water
Distribution System Security 7-18
7.6.1 TEVA Overview 7-18
7.6.1.1 Stochastic Modeling 7-19
7.6.1.2 Impact Analysis 7-19
7.6.1.3 Threat Mitigation Analysis 7-19
7.6.2 Application of TEVA to a Water Distribution System for Optimal Monitoring 7-19
7.6.2.1 Simulation Overview 7-20
7.6.2.2 TEVA Analysis Approach 7-20
7.6.2.3 TEVA Analysis Results 7-21
7.6.3 Case Summary 7-22
7.7 Field Testing of Water-Distribution Systems in Support of an Epidemiologic Study 7-22
7.7.1 Case Study Overview 7-22
7.7.2 Field Work 7-23
7.7.2.1 C-Factor and Fire-Flow Tests 7-23
7.7.2.2 Tracer Test and Continuous Measurements 7-23
7.7.3 Additional Test Procedures 7-24
7.7.4 Case Study Summary 7-25
7.8 Chapter Summary 7-25
References 7-25
viii
-------
A Reference Guide for Utilities
List of Tables
Table Page
Table 1-1. Selected Rules and Regulations Dealing with Distribution Systems (Not Inclusive) 1-9
Table 2-1. Available Hydraulic and Water Quality Network Modeling Software Packages 2-13
Table 2-2. Example CFD Modeling Software Packages 2-14
Table 2-3. Example Transient Modeling Software Packages 2-15
Table 3-1. Tracer Characteristics (adapted from Teefy, 1996) 3-7
Table 3-2. Equipment Costs 3-14
Table 3-3. Representative Labor Hours for a Range of Studies 3-15
Table 4-1 Calibration/Input Requirements for Water Quality Models 4-3
Table 4-2 Draft Calibration Criteria for Modeling (based on ECAC, 1999) 4-13
Table 5-1. Federal Distribution System Water Quality Monitoring Requirements 5-2
Table 7-1. Master Operating Criteria Used to Develop Operating Schedules for the Historical Water
Distribution System, Dover Township Area, NJ (from Maslia et al., 2001) 7-6
Table 7-2. Field Data Collected During Fire-Flow Test at Site H02 7-24
List of Figures
Figure Page
Figure 1-1. Water Treatment Process at the Miller Plant on the Ohio River (Adapted from: GCWW
2005) 1-1
Figure 1-2. Distribution System Interactions that Affect Water Quality (Adapted from: MSU,2005). . 1-3
Figure 1-3. Total Number and Proportion of U.S. Waterborne Diseases Associated with Water
Distribution System Deficiencies 1-6
Figure 1-4. Evolution of Federal Drinking Water Regulations -Timeline 1-8
Figure 2-1. Simple Link-Node Representation of a Water Distribution System 2-2
Figure 2-2. A Flow Chart for Estimating Future Water Demand Based on Land-Use Methodology. .. 2-5
Figure 2-3. EPANET Graphical Output Showing Flow and Pressure 2-7
Figure 2-4. Sample EPANET Time Series Plots of Flow, Pressure, and Tank Water Level 2-7
Figure 2-5. EPANET Sample Tabular Outputs (at time 10.00 hrs) 2-8
Figure 2-6. Illustration of the Evolution of Hydraulic and Water Quality Models 2-9
Figure 2-7. Graphical Output from a CFD Model Showing Tracer Concentration in a Tank 2-14
Figure 2-8. A Large Physical Model of a Tank (Source: Bureau of Reclamation Laboratory) 2-14
Figure 2-9. Graphical Output Based on 3-D Laser Induced Fluorescence with a Physical Scale
Model Showing Mixing in Tank (Source: Georgia Tech) 2-15
Figure 2-10. Negative Pressure Transient Associated with a Power Outage 2-16
Figure 2-11a. Velocity Field at a Junction 2-18
Figure 2-11b. Tracer Concentration at a Junction 2-18
Figure 3-1. Flow Calibration Tube 3-9
Figure 3-2. Automated Monitoring Station 3-10
Figure 3-3. Tracer Injection Setup (Storage Tank, Calibration Tube and Feed Pump) 3-11
Figure4-1. Conceptual Representation of Calibration 4-1
Figure 4-2. Effects of the Initial Water Age on the Modeled Results 4-3
Figure 4-3. Schematic of Standard Two-Gage C-Factor Test Setup 4-4
Figure 4-4. Schematic of Parallel Hose C-Factor Test Setup 4-4
Figure 4-5. Fire-Flow Test Setup 4-5
Figure 4-6. A Hydrant Being Flowed with a Diffuser as Part of a Fire-Flow Test 4-5
Figure 4-7. Schematic Representation of Small-Suburban Dead-End System 4-8
Figure 4-8. Empirical Relationship Between Chloride and Conductivity. 4-8
Figure 4-9. Sample Chloride Data Used at One Station for Calibration 4-9
Figure 4-10. Comparison of Model Versus Field Results for Continuous Monitor Location CM-18 at
Various Calibration Stages 4-9
ix
-------
A Reference Guide for Utilities
Figure4-11. Calibration of "Looped Portion." 4-10
Figure 4-12. Schematic Representation of Case 2 Study Location 4-11
Figure 4-13a. Modeled Flows Compared to Measured Flows Before Calibration 4-11
Figure 4-13b. Modeled Flows Compared to Measured Flows After Calibration 4-11
Figure 4-14a. Chloride Concentration for Calibration Event at Continuous Monitor Location CM-59. 4-12
Figure 4-14b. Chloride Concentration for Validation Event at Continuous Monitor Location CM-59... 4-12
Figure 5-1. Wastewater Injection: Free Chlorine and Associated Grab Sample Results 5-6
Figure 5-2. Theoretical Example of Benefits from Monitors 5-6
Figure 5-3. Pareto-Optimal Cost Effectiveness Diagram 5-7
Figure 5-4. Fort Reno #2 Remote Sampling System 5-12
Figure 5-5. WASA Remote Monitoring System Layout and Data Transmission Scheme 5-12
Figure 5-6. Monitoring Data for Fort Reno Tank 5-13
Figure 5-7. City of Tucson Water Quality Zone Map 5-13
Figure 5-8. Continuous Water Quality Monitoring Station 5-13
Figure 6-1. Mercator's Map of the World in 1569 (Whitfield, 1994) 6-2
Figure 6-2. Landsat Thematic Mapper™ Images of the Missouri River Floodplain Near Glasgow,
Missouri. (USGS, 1993) 6-4
Figure 6-3. Typical Inputs and Results of Current CIS Packages 6-5
Figure 6-4. Digital Terrain Model of Mount St. Helens after Eruption in 1980 (R. Home, 2004) 6-6
Figure 6-5. Map of Pressure Zone Showing Three Types of CIS Vector Data 6-6
Figure 6-6. Regional Land Cover Characterization as a Raster Database (USGS, 1992) 6-7
Figure 6-7. Triangulation of Elevation (Z) Data 6-7
Figure 6-8a. Typical Representation of a Pipe Section in CIS 6-11
Figure 6-8b. Typical Representation of a Pipe Section in a Network Model 6-11
Figure 6-9. LVVWD Distribution System Growth 6-12
Figure 6-10. Conceptual Relationship Model for Integration 6-12
Figure 6-11. One-to-One Relationship Between CIS and Network Modeling Data 6-12
Figure6-12. Pressure Complaint Resolution -CIS Parcel/Account Search Window. 6-13
Figure 6-13. Pressure Complaint Resolution - Parcel and Hydrant Location 6-14
Figure 6-14. Pressure Complaint Resolution - Model and Field Pressure Comparison 6-14
Figure 6-15. Water Main Break Analysis -Valve Isolation 6-15
Figure 6-16. Water Main Break Analysis-Impacted Customer List 6-15
Figure 6-17. Water Main Break Analysis-Comparison of Junction Pressures - Before and After
Shutdown 6-16
Figure 6-18. CIS Geo-coding - Metered Sales Demand Allocation Procedure 6-17
Figure 7-1. Comparison of Early Confirmed Cases of Salmonella Positive Sample Versus the
Estimated Distribution of Tank Water During the First 6 Hours of the Flushing Program. 7-3
Figure 7-2. Investigation Area, Dover Township, Ocean County, NJ (modified from Maslia et al.,
2001) 7-4
Figure 7-3. Distribution System Zones-Woburn, MA (May 1969) 7-5
Figure 7-4. Three-Dimensional Representation of Monthly Production of Water, Dover Township
Area, NJ (from Maslia etal., 2001) 7-6
Figure 7-5. Areal Distribution of Simulated Proportionate Contribution of Water from the Parkway
Wells (22, 23, 24, 26, 28, 29) to Locations in the Dover Township Area, NJ, July 1988
Conditions (from Maslia etal., 2001) 7-8
Figure 7-6. Simulated Proportionate Contribution of Water from Wells and Well Fields to Selected
Locations, Dover Township Area, NJ, July 1988 Conditions (from Maslia etal., 2001). ... 7-8
Figure 7-7. Estimated Upper 97.5 Percent Credibility Limit for Annual Perchlorate Intake by One
Plaintiff (Grayman, 2004) 7-9
Figure 7-8. Average Water Age in the Distribution System Over Last 24 Hours of a 2-Week
Simulation 7-11
Figure 7-9. Minimum Water Age in the Distribution System Over Last 24 Hours of a 2-Week
Simulation 7-11
Figure 7-10. Diurnal Water Age at Node J-456 7-12
Figure 7-11. Minimum Chlorine Residual in Distribution System Over Last 24 Hours of a 2-Week
Simulation 7-12
-------
A Reference Guide for Utilities
Figure 7-12. Zones Representing Potential Monitoring Locations for IDSE Based on Modeling.... 7-12
Figure 7-13. Water Distribution System Characteristics 7-13
Figure 7-14. Allowable Contaminant Introduction Locations 7-14
Figure 7-15. Contaminant Concentration Just Downstream of Contaminant Introduction Location
(Node 121) 7-14
Figure 7-16. Contaminant Concentration Far Downstream of Contaminant Introduction Location
(Node 143) 7-14
Figure 7-17. Monitoring Locations Selected by the Optimization Model 7-15
Figure 7-18. Hypothetical Water Distribution System Showing Pipelines 7-16
Figure 7-19. Conceptual Diagram Showing the Ranking and Prioritization Methodology. 7-16
Figure 7-20. Hypothetical System Showing High Score Areas (>27) Overlain with Hospitals and
Schools 7-17
Figure 7-21. Display of Low-Velocity Pipes, Oversized Pipes, and Current Monitoring Stations Using
the Spatial Database Display Tool 7-18
Figure 7-22. Threat Ensemble Vulnerability Assessment Framework 7-18
Figure 7-23. Comparison of 24-Hour, 48-Hour, and Real-Time, Continuous Contamination
Monitoring Systems for the Reduction in Mean Infections for a 24-Hour Contaminant
Attack 7-21
Figure 7-24. Comparison of 24-Hour, 48-Hour, and Real-Time, Continuous Contamination
Monitoring Systems for the Reduction in the Maximum Number of Infections for a 24-
Hour Contaminant Attack 7-21
Figure 7-25. Water Distribution Systems Serving U.S. Marine Corps Base, Camp Lejeune, NC.... 7-22
Figure 7-26. Continuous Recording Pressure Logger Mounted on Brass Shutoff Valve and Hydrant
Adapter Cap Used for Fire-Flow and C-Factor Tests 7-23
Figure 7-27. Location of Fire Hydrants Used in Fire-Flow Test at Site H02 7-23
Figure 7-28. Horiba W-23XD Dual Probe Ion Detector Inside Flow Cell 7-24
Figure 7-29. Arrival Times of the Calcium Chloride Tracer at Monitoring Locations in Hadnot Point
WTP Area, May 25, 2004 7-25
XI
-------
A Reference Guide for Utilities
Acronyms and Abbreviations
AC
ACCNSCM
ADAPT
ADEQ
AI2(S04)3
AM/FM
AMR
ASCE
ATSDR
AWWA
AwwaRF
C
CI04
CaCI2
CAD
CADD
CDC
CFD
CIS
CM
COGO
CRT
CWS
DBF
DBPR1
DBPR2
DC
D.C.
DEM
DIME
DLG
DSOP
DSS
DTM
Alternating Current
Arsenic and Clarifications to
Compliance and New Source
Contaminant Monitoring
Areal Design and Planning Tool
Arizona Department of
Environmental Quality
Aluminum Sulfate
Automated Mapping (or Asset
Management/Facilities
Management
Automated Meter Reading
American Society of Civil Engineers
Agency for Toxic Substances and
Disease Registry
American Water Works Association
Awwa Research Foundation
Coefficient of Roughness
Perchlorate Anion
Calcium Chloride
Computer-Aided Design
Computer-Aided Design and
Drafting
Centers for Disease Control and
Prevention
Computational Fluid Dynamics
Customer Information System
Continuous Monitoring
Coordinated Geometry
Cathode Ray Tube
Contamination Warning System
Disinfection By-Products
Disinfectant By-Product Rule -
Stage 1
Disinfectant By-Product Rule -
Stage 2
Direct Current
District of Columbia
Digital Elevation Model
Dual Independent Map Encoding
Digital Line Graph
Distribution System Water Quality
Optimization Plan
Distribution System Simulator
Digital Terrain Model
DWQM Dynamic Water Quality Model
EBMUD East Bay Municipal Utility District
EDM Electronic Distance Measurement
EMPACT Environmental Monitoring for Public
Access and Community Tracking
EOSAT Earth Observation Satellite
EPA U.S. Environmental Protection
Agency
EPS Extended Period Simulation
ESSA Environmental Science Services
Administration
EWS Environmental Warning System
FC Fecal Coliform
FeCI3 Ferric Chloride
FOH Federal Occupational Health
GA Genetic Algorithm
gal Gallon
GBF Geographic Base File
GC Gas Chromatograph
GCWW Greater Cincinnati Waterworks
CIS Geographic Information System
GPD Gallons Per Day
gpm Gallons Per Minute
GPS Global Positioning System
GRASS Geographic Resources Analysis
Support System
GUI Graphical User Interface
GWR Ground Water Rule
HAA HaloaceticAcid
HAA5 The five Haloacetic Acids
HACCP Hazard Analysis Critical Control
Point
HGL Hydraulic Grade Line
HSPP Health and Safety Project Plan
ICR Information Collection Rule
IDSE Initial Distribution System
Evaluation
IESWTR Interim Enhanced Surface Water
Treatment Rule
ILSI International Life Sciences Institute
I/O Input/Output
ISE Ion Selective Electrode
ISO Insurance Services Office
LCR Lead and Copper Rule
XII
-------
A Reference Guide for Utilities
LIFO Last In/First Out
LIMS Laboratory Information
Management System
LIS Land Information System
LT1ESWTR Long Term 1 Enhanced Surface
Water Treatment Rule
LT2ESWTR Long Term 2 Enhanced Surface
Water Treatment Rule
LVVWD Las Vegas Valley Water District
ma Milli-Amperes
MCL Maximum Contaminant Level
MCLG Maximum Contaminant Level Goal
MDL Minimum Detection Limit
MDNR Missouri Department of Natural
Resources
MDOH Missouri Department of Health
MGD Million Gallons per Day
mg/L milligrams per liter
MIT Massachusetts Institute of
Technology
MOC Master Operating Criteria
MRDLG Maximum Residual Disinfectant
Level Goals
MS Mass Spectrometer
MSU Montana State University
NaCI Sodium Chloride
NAD27 North American Datum of 1927
NAD83 North American Datum of 1983
NAPP National Aerial Photography
Program
NASA National Aeronautics and Space
Administration
NFRA National Fire Protection Association
NHAP National High Altitude Photography
NIPDWR National Interim Primary Drinking
Water Regulations
NJDHSS New Jersey Department of Health
and Senior Services
NMWD North Marin Water District
NPL National Priorities List
NPWA North Penn Water Authority
NOM Naturally Occurring Organic (and/
or Inorganic) Matter
NRC National Research Council
O&M Operations and Maintenance
OCMS Online Contaminant Monitoring
System
ODBC Open Database Connectivity
ORD Office of Research and Development
ORP Oxidation Reduction Potential
PAB3D A Three-Dimensional Computational
Fluid Dynamics Model developed by
Analytical Services & Materials, Inc.
PC Personal Computer
PDD Presidential Decision Directive
PHRP Public Health Response Plan
PL Public Law
POE Point of Entry
POGA Progressive Optimality Genetic
Algorithm
psi Pounds Per Square Inch
PVC Polyvinyl Chloride
PWS Public Water System
QA Quality Assurance
QAPP Quality Assurance Project Plan
QC Quality Control
RDBMS Relational Database Management
Systems
RDWR Radon in Drinking Water Rule
SAN Styrene Acrylonitrile
SCADA Supervisory Control and Data
Acquisition
SCCRWA South Central Connecticut Regional
Water Authority
SDMS Spatial Database Management System
SDWA Safe Drinking Water Act
SDWAA Safe Drinking Water Act Amendments
SMP Standard Monitoring Program
SNL Supply Node Link
SOP Standard Operating Procedure
SPC State Plane Coordinates
SSS System Specific Study
SVOC Semivolatile Organic Compound
SWTR Surface Water Treatment Rule
SYMAP Synagraphic Mapping
T&E Test and Evaluation
TCE Trichloroethylene
TCR Total Coliform Rule
TEVA Threat Ensemble Vulnerability
Assessment
THM Trihalomethane
TIGER Topologically Integrated Geographic
Encoding and Referencing
XIII
-------
A Reference Guide for Utilities
TIN Triangulated Irregular Network
TIROS1 Television and Infrared Observation
Satellite 1
TOC Total Organic Carbon
TT Treatment Technique
TTHM Total Trihalomethane
TV Television
U.S. United States
UF Ultrafiltration
UHF Ultra High Frequency
UV Ultraviolet
UV-Vis Ultraviolet-Visible
USGS United States Geological Survey
UTM Universal Transverse Mercator
VHF Very High Frequency
WASA Water and Sewer Authority
WATERS Water Awareness Technology
Evaluation Research and Security
WOP Water Quality Parameter
WRC Water Research Centre
WSSM Water Supply Simulation Model
WSTP Wells, Storage Tanks, and Pumps
WTP Water Treatment Plant
XIV
-------
A Reference Guide for Utilities
Acknowledgments
EPA would like to acknowledge the principal authors of the document, "Water Distribution System
Analysis: Field Studies, Modeling, and Management," a Reference Guide for Utilities. The authors
were Mr. Srinivas Panguluri, P.E. (Shaw Environmental, Inc.), Dr. Walter M. Grayman, P.E. (WM.
Grayman Consulting Engineer), and Dr. Robert M. Clark, P.E., D.E.E. (Environmental Engineering and
Public Health Consultant). The work was performed under EPA Contract No. EP-C-04-034, Work
Assignment No. 0-08 and 1-08, with Shaw Environmental, Inc.
EPA acknowledges the significant contributions from Mr. David J. Hartman and Dr. Yeongho Lee,
P.E., of the Greater Cincinnati Water Works. Mr. Hartman and Dr. Lee served as external reviewers
and directly contributed to the development of portions of this document. Greater Cincinnati Water
Works also participated significantly in the field tracer tests that served as a foundation for the
development of this document.
EPA acknowledges the peer reviews by the following utility personnel: Mr. William R. Kirkpatrick and
Mr. Ronald B. Hunsinger of East Bay Municipal Utility District, and Mr. Arnold Strasser, P.E., of Denver
Water.
EPA also acknowledges the peer reviews provided for this document by Dr. Fran A. DiGiano, P.E., of
University of North Carolina, Mr. Morris L. Maslia, P.E., D.E.E., of the Agency for Toxic Substances
and Disease Registry, Dr. Edward A. McBean, P.E., of University of Guelph, Dr. Vanessa L. Speight,
P.E., of Malcolm Pirnie, Ms. L. Michelle Moore of National Drinking Water Clearinghouse, and Mr. E.
Radha Krishnan, P.E., of Shaw Environmental, Inc.
EPA also gratefully acknowledges the following contributors to this document: Mr. Daniel R. Quintanar
of Tucson Water, Ms. Laura B. Jacobson, P.E., Mr. Sridhar Kamojjala, P.E., and Mr. Mao Fang, P.E., of
Las Vegas Valley Water District, Mr. Morris L. Maslia, P.E., D.E.E., of the Agency for Toxic Substances
and Disease Registry, Dr. William B. Samuels, P.E., of SAIC, Dr. Mark W LeChevallier and Dr. Kala K.
Fleming of American Water.
EPA Contributors
Ms. Shirley J. Gibson functioned as Project Officer of EPA Contract No. EP-C-04-034, Ms. Lucille M.
Garner served as Work Assignment Manager and Mr. Craig L. Patterson, P.E., served as the Alternate
Work Assignment Manager and technical reviewer for this document. Mr. Roy C. Haught served as
EPA Technical Advisor. Mr. Jonathan G. Herrmann, P.E., Mr. Kim R. Fox, P.E., D.E.E., and Dr. James
A. Goodrich performed technical reviews of the document. Mr. Stephen M. Harmon was the EPA
Quality Assurance Manager, and was responsible for the quality assurance review of the document.
The document was also reviewed by many other EPA reviewers including the following: Mr. Blake L.
Atkins, Mr. William H. Davis, Ms. Jean E. Dye, Mr. Robert J. Janke, Mr. Bruce A. Macler, Ms. Jill R.
Neal, Dr. Lewis A. Rossman, P.E., Mr. Kenneth H. Rotert, Dr. Irwin J. Silverstein, P.E., D.E.E., and Ms.
Elin A. Warn.
xv
-------
A Reference Guide for Utilities
Chapter 1
Introduction
Drinking water utilities in the United States (U.S.)
and throughout the world face the challenge of
providing water of good quality to their consumers.
Frequently, the water supply is derived from surface
water or groundwater sources that may be subject to
naturally occurring or accidentally introduced
contamination (ILSI, 1999; Gullick et al., 2003). In
other cases, routine upstream waste discharges or
purposeful contamination of the water can diminish
the quality of the water. The treated water may be
transmitted through a network of corroded or
deteriorating pipes. All of these factors can result
in degradation in the quality of the water delivered
to customers.
In the U.S., drinking water quality has to comply
with federal, state, and local regulations. This is
based on selected physical, chemical, and biologi-
cal characteristics of the water. The U.S. Environ-
mental Protection Agency (EPA) has promulgated
many drinking water standards under the Safe
Drinking Water Act (SDWA) of 1974. These rules
and regulations require that public water systems
(PWSs) meet specific guidelines and/or numeric
standards for water quality. The SDWA defines a
PWS as a system that serves piped water to at least
25 persons or 15 service connections for at least 60
days each year. For the purposes of this reference
guide, PWSs are referred to as utilities.
The SDWA has established two types of numeric
standards. The first type of standard is enforceable
and referred to as a maximum contaminant level
(MCL). The other non-enforceable standard is
referred to as a maximum contaminant level goal
(MCLG). MCLGs are set at a level at which no known
or anticipated adverse human health effects occur.
Where it is not economically or technologically
feasible to determine the level of a contaminant, a
treatment technique (TT) is prescribed by EPA in lieu
of establishing an MCL. For example, Giardia is a
microbial contaminant that is difficult to measure. To
ensure proper removal, experimental work has
established optimum treatment conditions for the
water at a specified pH, temperature, and chlorine
concentration for a specified length of time to achieve
a fixed level of inactivation.
Compliance with MCL and TT requirements is
typically ensured by requiring that water utilities
periodically monitor various characteristics of the
treated water. In summary, the EPA Guidelines and
Standards are designed to ensure that drinking
water is adequately treated and managed by water
Removing contaminants from drinking water can be
expensive. Depending upon the type and level of
contaminant(s) present in the source water, utilities can
choose from a variety of treatment processes. These
individual processes can be arranged in a "treatment
train" (a series of processes applied in a sequence).
The most commonly used treatment processes include
coagulation/flocculation, sedimentation, filtration, and
disinfection. Some water systems also use ion ex-
change, membrane separation, ozonation, or carbon
adsorption for treatment. The basic treatment options
are briefly discussed later in this chapter. As an
example, Figure 1-1 depicts the water treatment
process implemented by the Greater Cincinnati Water
Works (GCWW) at the Miller Plant on the Ohio River.
Presetting
removes most solids
Final settling occurs,
water prepared
Further settling for final treatment
occurs in reservoir
chlorine added,
fluoride added
Reservoir
To distribution Hj
system
~*^B Furnace ->•
cleans carbon
for reuse
Figure 7-7. Water Treatment Process at the Miller
Plant on the Ohio River (Adapted from: GCWW 2005).
utilities to support public safety, protect public
health, and promote economic growth (Clark and
Feige, 1993).
Disinfection of drinking water is considered to be one
of the major public health advances of the 20th
century. The successful application of chlorine as a
disinfectant was first demonstrated in England. In
1908, Jersey City (NJ) initiated the use of chlorine for
water disinfection in the U.S. This approach subse-
quently spread to other locations, and soon the rates
of common epidemics such as typhoid and cholera
dropped dramatically. Today, disinfection is an
essential part of a drinking water treatment train.
Chlorine, chlorine dioxide, and chloramines are most
1-1
-------
A Reference Guide for Utilities
While disinfectants are effective in controlling many
microorganisms, they can react with naturally occurring
organic (and/or inorganic) matter (NOM) in the treated
and/or distributed water to form potentially harmful
disinfection byproducts (DBFs). Many of these DBFs
are suspected of causing cancer, reproductive, and
developmental problems in humans. To minimize the
formation of DBFs, EPA has promulgated regulations
that specify maximum residual disinfectant level goals
(MRDLGs) for chlorine (4 milligrams per liter [mg/L] as
chlorine), chloramines (4 mg/L as chlorine), and
chlorine dioxide (0.8 mg/L as chlorine dioxide). In
addition, MCLs for the DBFs total trihalomethanes
(TTHMs) and haloacetic acids (HAAS) have been
established as 0.080 and 0.060 mg/L, respectively. The
TTHMs include chloroform, bromodichloromethane,
dibromochloromethane and bromoform. The HAAS
include monochloroacetic acid, dichloroacetic acid,
trichloroacetic acid, monobromoacetic acid and
dibromoacetic acid. In order to meet these require-
ments, utilities may need to remove the DBF precursor
material from the water prior to disinfection by apply-
ing appropriate treatment techniques or modify their
disinfection process.
often used because they are very effective disinfec-
tants, and residual concentrations can be maintained
in the water distribution system. Some utilities (in the
U.S. and Europe) use ozone and chlorine dioxide as
oxidizing agents for primary disinfection prior to the
addition of chlorine or chlorine dioxide for residual
disinfection. The Netherlands identifies ozone as the
primary disinfectant, as well as common use of
chlorine dioxide, but typically uses no chlorine or
other disinfectant residual in the distribution system
(Cornell, 1998).
Prior to the passage of the SDWA of 1974, most
Some important distribution system water quality
concerns are: maintenance of proper disinfectant levels;
minimization of DBF formation; turbidity, taste, color,
and odor issues; distribution tank mixing and utiliza-
tion; main repair and pressure stabilization; flow
management; cross-connection control and back-flow
prevention.
Some water quality goals are contradictory. For
example, an important goal is to maintain a positive
disinfectant residual in order to protect against micro-
bial contamination. However, DBFs (TTHMs) will
increase as water moves through the network as long as
disinfectant residual and NOM is available. Other DBFs
(HAAS) are degraded biologically when free chlorine or
chloramines are nearly absent.
drinking water utilities focused on meeting drink-
ing water standards at the treatment plant, even
though it had long been recognized that water
quality can deteriorate in a distribution system.
The SDWA introduced a number of MCLs that must
be measured at various monitoring points in the
distribution system. Consequently, water quality in
the distribution system became a focus of regula-
tory action and of major interest to drinking water
utilities. Subsequently, utilities worked with
various research organizations (including EPA) to
understand the impact of the distribution system on
water quality. The collective knowledge from this
research has been applied to improve the quality of
water delivered to the consumer (Clark and
Grayman, 1998).
Prior to September 11, 2001 (9/11), few water utilities
were using online monitors in a distribution system as a
means of ensuring that water quality was being main-
tained and addressed in cases of deviation from estab-
lished ranges. Now the enhanced focus on water
security has led EPA and water utilities to collectively
evaluate commercial technologies to remotely monitor
the distribution system water quality in real-time. As a
part of an evolutionary process, in the future, these
monitoring technologies are expected to be integrated
with computer modeling and geospatial technologies.
This evolution of monitoring and modeling technolo-
gies can potentially minimize the risks from drinking
water contaminants in distribution systems.
This reference guide has been prepared to provide
information to drinking water utilities and research-
ers on the state of the art for distribution system
management and modeling. Guidance is provided
on the application of advanced modeling tools that
can enhance a utility's ability to better manage
distribution system water quality. This introduc-
tory chapter provides the basic concepts, which
include:
• Distribution system - infrastructure design and
operation (definitions and overview).
• Water quality problems and issues (a brief
review).
• Regulatory framework (an overview).
• Assessment and management of water quality
(current practices).
• Advanced tools for water quality management
(in distribution systems).
Subsequent chapters will provide more details on
related concepts and tools.
1-2
-------
A Reference Guide for Utilities
1.1 Distribution System -
Infrastructure Design and
Operation
Distribution system infrastructure is a major asset of a
water utility, even though most of the components are
either buried or located inconspicuously. Drinking
water distribution systems are designed to deliver
water from a source (usually a treatment facility) in
the required quantity, quality, and at satisfactory
pressure to individual consumers in a utility's service
area. In general, to continuously and reliably move
water between a source and a customer, the system
would require storage reservoirs/tanks, and a network
of pipes, pumps, valves, and other appurtenances.
This infrastructure is collectively referred to as the
drinking water distribution system (Walski et al., 2003).
1.1.1 Key Infrastructure Components
A detailed description of the various distribution
system infrastructure components is readily available
from other sources and beyond the scope of this
document. However, for the purposes of establishing
the basics, this section includes a brief discussion of
the uses of the major components, their characteris-
tics, general maintenance requirements, and desirable
features.
1.1.1.1 Storage Tanks/Reservoirs
Tanks and reservoirs are used to provide storage
capacity to meet fluctuations in demand, to provide
reserves for fire-fighting use and other emergency
situations, and to equalize pressures in the distribu-
tion system. The most frequently used type of storage
facility is the elevated tank, but other types of tanks
and reservoirs include in-ground tanks and open or
closed reservoirs. Materials of construction include
concrete and steel. An issue that has drawn a great
deal of interest is the problem of water turnover
within storage facilities. Much of the water volume in
storage tanks is dedicated to fire protection. Unless
utilities make a deliberate effort to exercise (fill and
draw) their tanks, or to downsize the tanks when the
opportunity presents itself, there can be both water
aging and water mixing problems. The latter can lead
to stratification and/or large stagnant zones within the
water volume. Some of these issues will be discussed
later in this document.
1.1.1.2 Pipe Network
The system of pipes or "mains" that carry water from
the source (such as a treatment plant) to the consumer
is often categorized as transmission/trunk, distribu-
tion, and service mains. Transmission/trunk mains
usually convey large amounts of water over long
distances, such as from a treatment facility to a
storage tank within the distribution system. Distribu-
tion mains are typically smaller in diameter than the
transmission mains and generally follow city streets.
Service mains are pipes that carry water from the
distribution main to the building or property being
served. Service lines can be of any size, depending
on how much water is required to serve a particular
customer, and are sized so that the utility's design
pressure is maintained at the customer's property for
the desired flows. The most commonly used pipes
today for water mains are ductile iron, pre-stressed
concrete, polyvinyl chloride (PVC), reinforced
plastic, and steel. In the past, unlined cast iron pipe
and asbestos-cement pipes were frequently used.
Even a medium-sized water utility may have thou-
sands of miles of pipes constructed from various types
of materials, ranging from new, lined or plastic pipes
to unlined pipes that are more than 50 years old. Over
time, biofilms and tubercles attached to pipe walls can
result in both loss of carrying capacity and a significant
loss of disinfectant residual, thereby adversely
affecting water quality (Clark and Tippen, 1990).
Figure 1-2 depicts the various distribution system
interactions that may adversely affect water quality.
The Distribution System as Reactor
PIPE SURFACE
Corrosion
Surface (Chemical & Biological)
Reactions
» »-'Bulk
k"ReactiA= organisms
Participates Heterotrophs Coliforms
Biofilm/regrowth
Figure 1-2. Distribution System Interactions that Affect
Water Quality (Adapted from: MSU, 2005).
The mains should be placed in areas along the public
right of way, which provides for ease of access,
installation, repair, and maintenance. Broken or
leaking water mains should be repaired as soon as
possible to minimize property damage and loss of
water. In the past, it has been standard practice to
maintain the carrying capacity of the pipe in the
distribution system as high as possible to provide the
design flow and keep pumping costs as low as
possible. However, there has been recent concern that
excess capacity can lead to long residence times and
thus contribute to deterioration in water quality.
1.1.1.3 Valves
There are two general types of valves in a distribution
system: isolation valves and control valves. Isolation
valves are used in the distribution system to isolate
sections for maintenance and repair and are typically
1-3
-------
A Reference Guide for Utilities
located in a system so that the areas isolated will
cause a minimum of inconvenience to other service
areas. Maintenance of the valves is one of the major
activities carried out by a utility. Many utilities have
a regular valve-turning program in which a percentage
of the valves are opened and closed on a regular basis.
It is desirable to turn each valve in the system at least
once per year. In large systems, this may or may not
be practical, but periodic exercise and checking of
valve operations should occur. This practice mini-
mizes the likelihood that valves will become inoper-
able due to corrosion. The implementation of such a
program ensures that, especially during an emergency.
water can be shut off or diverted and that valves have
not been inadvertently closed.
Control valves are used to regulate the flow or
pressure in a distribution system. Typical types of
control valves include pressure-reducing valves.
pressure-sustaining valves, flow-rate control valves.
throttling valves, and check valves.
1.1.1.4 Pumps
Pumps are used to impart energy to the water in order
to boost it to higher elevations or to increase pressure.
Routine maintenance, proper design and operation.
and testing are required to insure that they will meet
their specific objectives. Pump tests are typically run
every five to ten years to check the head-discharge
relationship for the pump. Many system designers
recommend two smaller pumps instead of one large
pump to ensure redundancy.
1.1.1.5 Hydrants and Other Appurtenances
Hydrants are primarily a part of the fire-fighting
infrastructure of a water system. Although water
utilities usually have no legal responsibility for fire
flow, developmental requirements often include fire
flows, and thus, distribution systems are designed to
support needed fire flows where practical (AWWA.
1998). Proper design, spacing, and maintenance are
needed to insure an adequate flow to satisfy fire-
fighting requirements. Fire hydrants are typically
exercised and tested periodically by water utility or
fire department personnel. Fire-flow tests are con-
ducted periodically to satisfy the requirements of the
Insurance Services Office (ISO, 2003) or as part of a
water distribution system calibration program. Other
appurtenances in a water distribution system include
blow-off valves and air release valves.
1.1.2 Basic Design and Operation Philosophy
A detailed understanding of "how water is used" is
critical to understanding water distribution system
design and operation. Almost universally, the manner
in which industrial and residential customers use
water drives the overall design and operation of a
water distribution system. Generally, water use varies
Conservative design philosophies, aging water supply
infrastructure, and increasingly stringent drinking water
standards have resulted in concerns over the viability of
drinking water systems in the U.S. Questions have been
raised over the structural integrity of these systems as
well as their ability to maintain water quality from the
treatment plant to the consumer. The Clean Water and
Drinking Water Infrastructure Gap Analysis (EPA 2002),
which identified potential funding gaps between
projected needs and spending from 2000 through 2019,
estimated a potential 20-year funding gap for drinking
water capital, and operations and maintenance, ranging
from $45 billion to $263 billion, depending on spend-
ing levels. Based on current spending levels, the U.S.
faces a shortfall of $11 billion annually to replace aging
facilities and comply with safe drinking water regula-
tions. Federal funding for drinking water in 2005
remained level at $850 million—less than 10% of the
total national requirement (ASCE, 2005). Parts of many
systems are approaching or exceed 100 years old, and an
estimated 26 percent of the distribution system pipe in
this country is unlined cast iron and steel in poor
condition. At current replacement rates for distribution
system components, it is projected that a utility will
replace a pipe every 200 years (Kirmeyer et al., 1994).
Grigg, NS, 2005, provides comprehensive guidance to
utilities on how to assess options for distribution system
renewal. Grigg's report contains a knowledge base on
condition assessment, planning and prioritization, and
renewal methods.
both spatially and temporally. Besides customer
consumption, a major function of most distribution
systems is to provide adequate standby fire-flow
capacity (Fair and Geyer, 1971). For this purpose, fire
hydrants are installed in areas that are easily acces-
sible by fire fighters and are not obstacles to pedestri-
ans and vehicles. The ready-to-serve requirements for
fire fighting are governed by the National Fire
Protection Association (NFPA) that establishes
standards for fire-fighting capacity of distribution
systems (NFPA, 2003). In order to satisfy this need for
adequate standby capacity and pressure (as mentioned
earlier), most distribution systems use standpipes.
elevated tanks, and large storage reservoirs. Addi-
tionally, most distribution systems are "zoned."
Zones are areas or sections of a distribution system of
relatively constant elevation. Zones can be used to
maintain relatively constant pressures in the system
over a range of ground elevations. Sometimes, zone
development occurs as a result of the manner in which
the system has expanded.
The effect of designing and operating a system to
maintain adequate fire flow and redundant capacity
can result in long travel times for water between the
1-4
-------
A Reference Guide for Utilities
Non-potable waters (e.g., sea, river, and lake water)
without adequate treatment have been used for fire
protection for many years, often with disastrous results.
However, reclaimed wastewater (in cases where its
quality is better managed than the aforementioned
unregulated sources) has been effectively used for
providing fire protection (AwwaRF, 2002). St. Peters-
burg, FL, has been operating such a system to bolster
fire-protection capacity since 1976. The reclaimed
water hydrants are distinguished from potable water
hydrants by color and their special valves. If the
reclaimed water system is designed for fire protection,
the potable water piping can have a very small diameter
and investments can be made in higher quality pipe
materials, which, with much shorter residence time in
the system, would vastly improve the quality of the
water at the tap. With this in mind, where retrofitting
one of the two systems is necessary, it might be wiser to
use the existing potable water system for the reclaimed
water and retrofit with new, high-quality, smaller,
potable water lines (Okun, D., 1996).
treatment plant and the consumer. These long travel
times and low velocities may be detrimental to
meeting the drinking water MCLs. Long residence
times may lead to formation of DBFs, loss of disinfectant
residuals, bacterial growth, and formation of biofilm.
1.1.2.1 Pipe-Network Configurations
The branch and grid/loop are the two basic configura-
tions for most water distribution systems. A branch
system is similar to that of a tree branch with smaller
pipes branching off larger pipes throughout the
service area. This type of system is most frequently
used in rural areas, and the water has only one
possible pathway from the source to the consumer. A
grid/loop system is the most widely used configura-
tion in large municipal systems and consists of
interconnected pipe loops throughout the area to be
served. In this type of system, there are several
pathways that the water can follow from the source to
the consumer. Transmission mains are typically 20 to
24 inches in diameter or larger. Dual-service mains
that serve both transmission and distribution purposes
are normally 12 to 20 inches in diameter. Distribu-
tion mains are usually 6 to 12 inches in diameter in
every street. Service lines are typically 1 inch in
diameter. Specific pipe sizes can vary depending on
the extent of the distribution system and the magni-
tude of demand. Looped systems provide a high
degree of reliability should a line break occur,
because the break can be isolated with little impact
on consumers outside the immediate area (Clark and
Tippen, 1990; Clark et al., 2004).
1.1.2.2 Multiple Source Configuration
Many systems serve communities with multiple
sources of supply, such as a combination of wells and/or
surface sources. In a grid/looped system, this configu-
ration will influence water quality in a distribution
system due to the effect of mixing of water from these
different sources. These interactions are a function of
complex system hydraulics (Clark et al., 1988; Clark
et al., 1991a). Water quality models can be very
useful in defining mixing and blending zones within
water utility distribution networks. Mixing of water
in a network can result in taste and odor problems or
other water quality problems and can influence
maintenance, repair, and rehabilitation procedures.
1.1.2.3 Impact of System Design and Operation
on Water Quality
Based on the design and configuration of a particular
system, there are many opportunities for water quality
to change as water moves between the treatment plant
and the consumer. These unwanted changes may
occur due to various reasons including: failures at the
treatment barrier, transformations in the bulk phase.
corrosion and leaching of pipe material, biofilm
formation, and mixing between different sources of
water. Many researchers have investigated the factors
that influence water quality deterioration once it
enters the distribution system. It has been well
documented that bacteriological growth can cause
taste-and-odor problems, discoloration, slime buildup.
and economic problems, including corrosion of pipes
and bio-deterioration of materials (Water Research
Centre, 1976). Bacterial numbers tend to increase
during distribution and are influenced by several
factors, including bacterial quality of the finished
water entering the system, temperature, residence
time, presence or absence of a disinfectant residual.
construction materials, and availability of nutrients
for growth (Geldreich et al., 1972; LeChevallier et al..
1987; Maul et al., 1985a and b; Zhang and DiGiano,
2002; Camper et al., 2003).
It is difficult and expensive to study the problems
caused by system design and configuration in full-
scale systems. For example, one approach to
studying residual chlorine levels in dead-end or
low-flow situations would be to construct a pilot-
scale pipe system to simulate the phenomena.
Another approach would be to use mathematical
hydraulic and water quality models for simulation.
For either of these approaches to work, they must
be properly configured and/or calibrated to closely
simulate a full-scale system. A combination of
these approaches may be used to assess various
operational and design decisions, to determine the
impacts resulting from the inadvertent or deliberate
introduction of a contaminant into the distribution
system, and to assist in the design of systems to
improve water quality.
1-5
-------
A Reference Guide for Utilities
In pipes, it has been found that chlorine can be lost
through both the interaction with NOM in the bulk
phase and with pipe walls themselves in transporting
finished water. This mechanism for loss of chlorine may
be even more serious than long residence times in
tanks. The pipe wall demand, possibly due to biofilm
and tubercles, may use up the chlorine very rapidly in a
distribution system. Maintaining adequate levels of
disinfectant residual may require routine cleaning/
replacement of pipes and intensive treatment (Clark et
al., 1993a).
1.2 Water Quality Problems and
Issues
Drinking water treatment in the U. S. has played a
major role in reducing waterborne disease. For
example, the typhoid death rate for a particular year
in the 1880s was 158 per 100,000 in Pittsburgh, PA,
compared with 5 per 100,000 in 1935. Such dramatic
reductions in waterborne disease outbreaks were
brought about by the application of drinking water
standards and engineering "multiple barriers" of
protection. The multiple-barrier concept includes the
use of conventional treatment (e.g., sand filtration) in
combination with disinfection to provide safe and
aesthetically acceptable drinking water. The residual
disinfectant levels served to protect the water quality
within the distribution system prior to its delivery to
the consumer (Clark et al., 1991b).
Despite the passage of the SDWA, waterborne out-
breaks still occur. Two extensively studied examples
of waterborne disease in the U.S. were an Escherichia
coli O157:H7 (E. coli) outbreak in Cabool, Missouri.
in 1989 and a Salmonella outbreak in Gideon.
Missouri, in 1993. These two examples, discussed
later in Chapter 7, illustrate the importance of the
multiple-barrier concept. In both cases, the water
source was un-disinfected groundwater and the
utility's infrastructure was breached, allowing
contaminants to enter the system. This contamination
resulted in major waterborne outbreaks. Water quality
modeling was used in both cases to identify the
source of the outbreaks and to study the propagation
of the outbreak through the distribution network
(Clark etal., 1993aandb).
One useful outcome of the outbreaks in Missouri is
that the ensuing investigative studies have typically
led to the development and enhancement of scientific
analysis techniques. For example, the Gideon
Salmonella outbreak conclusions were based on
statistical studies performed by Centers for Disease
Control and Prevention (CDC) and corroborated by
water quality modeling performed by EPA. The study
provides an example of how tools such as water
quality models can be used to reliably study contami-
nant propagation in a distribution system (Clark et al..
1996). Both the Gideon and Cabool incidents were
associated with source water contamination, inad-
equate treatment, and breeches in the distribution
system.
These types of problems are not just isolated incidents
of infrastructure breakdowns. In fact, several prob-
lems with drinking water systems in the U. S. have
been identified by researchers. The National Research
Council (NRC, 2005) examined the causes of water-
borne outbreaks reported by various investigators
between 1971 and 2004. Figure 1-3 presents the total
number and proportion of waterborne diseases
associated with distribution system deficiencies
100
~ I?
Total Number of Outbreaks
Number of Outbreaks Due to Distribution System Deficiency
Percent Due to Distribution System Deficiency
nii
1971-1974 1975-1978 1979-1982 1983-1986 1987-1990 1991-1994 1995-1998 1999-2000 2001-2002
Figure 1-3. Total Number and Proportion of U.S. Waterborne Diseases Associated with Water Distribution
System Deficiencies.
1-6
-------
A Reference Guide for Utilities
On December 16, 1974, the U. S. Congress passed the
SDWA, which authorized the EPA to promulgate
regulations which would "protect health to the extent
feasible, using technology, treatment techniques, and
other means, which the Administrator determines are
generally available (taking costs into considera-
tion)... "(SDWA, 1974). As a result, a set of regulations
was promulgated in 1975 which became effective June
24, 1977. These were known as the National Interim
Primary Drinking Water Regulations (NIPDWR). The
NIPDWR established MCLs for 10 inorganic contami-
nants, six organic contaminants, turbidity, coliform,
radium-226, radium-228, gross alpha activity, and
man-made radionuclides. The NIPDWR also estab-
lished monitoring and analytical requirements for
determining compliance.
(extracted from the NRC report). As the figure
reveals, overall there is a general decrease in the
total number of waterborne disease outbreaks
during the reported period. However, there is a
general increase in the percentage of outbreaks that
are associated with distribution system deficiencies.
The NRC report attributes this increase in percent-
age of outbreaks (attributable to distribution system
deficiencies) to the lack of historical regulatory
focus on distribution systems.
1.3 Regulatory Framework
Concerns about waterborne disease and uncon-
trolled water pollution resulted in federal water
quality legislation starting in 1893 with the
passage of the Interstate Quarantine Act and
continuing to 1970 under the stewardship of the
U.S. Public Health Service (AWWA, 1999). Even
though significant advances were made to eliminate
waterborne disease outbreaks during that period.
the focus of drinking water concerns began to
change with the formation of the EPA in late 1970.
By the 1970s, more than 12,000 chemical com-
pounds were known to be in commercial use and
many more were being added each year. Many of
these chemicals cause contamination of groundwa-
ter and surface water, and are known to be carcino-
genic and/or toxic. The passage of the SDWA of
1974 was a reflection of concerns about chemical
contamination. In this section, a brief overview of
the regulatory framework is presented. A detailed
history of the evolution of the federal drinking
water regulations is beyond the scope of this
document.
Early in the history of the SDWA, the major focus of
EPA was to implement the Act and to initiate the
regulatory process. The first MCL established
under the SDWA was the TTHM Rule in 1979.
However, after several years of developing regula-
tions, it became obvious that the rulemaking
process must extend beyond a focus on MCLs at the
treatment plant and into the distribution system.
Many water utilities in the U.S. using surface
supplies were experiencing waterborne outbreaks.
especially from Giardia. The 1986 SDWA Amend-
ments laid the groundwork for the promulgation of
the Total Coliform Rule (TCR) and the Surface
Water Treatment Rule (SWTR) in 1989. The 1986
SDWA Amendments also set forth an aggressive
plan to eliminate lead from PWSs and resulted in
the promulgation of the Lead and Copper Rule
(LCR) in 1991. These actions therefore extended
the SDWA beyond its focus on the treatment plant
and into the distribution system (Owens, 2001).
A summary of the evolution of federal drinking water
regulation since the passage of the SDWA in 1974 is
presented in Figure 1-4. In addition to the rules and
regulations promulgated under the SDWA, security
has recently become an issue for the water utility
industry. Security of water systems is not a new issue.
The potential for natural, accidental, and purposeful
contamination of water supplies has been the subject
of many studies. For example, in May 1998, President
Clinton issued Presidential Decision Directive (FDD)
63 that outlined a policy on critical infrastructure
protection, including our nation's water supplies.
However, it was not until after September 11, 2001.
that the water industry focused on the vulnerability of
the nation's water supplies to security threats. In
recognition of these issues, President George W Bush
signed the Public Health Security and Bioterrorism
Preparedness and Response Act of 2002 (Bioterrorism
Act) into law in June 2002 (PL107-188). Under the
requirements of the Bioterrorism Act, community
water systems (CWSs) serving more than 3,300 people
are required to prepare vulnerability assessments and
emergency response plans. CWSs are PWSs that
supply water to the same population throughout the
year.
Table 1-1 summarizes the key requirements of the
regulations presented in Figure 1-4 from a distribu-
tion system compliance perspective.
Many of the tools and techniques discussed in this
reference guide can assist in complying with the
rules and regulations and security issues discussed
above. Water quality modeling techniques can be
used to identify points in the distribution system
that experience long retention times, which can in
turn represent locations in the system that may
experience chlorine residual loss, excessive
formation of DBFs, and the formation of biofilms.
Chlorine residual loss, in conjunction with biofilm
1-7
-------
A Reference Guide for Utilities
SDWA
Safe Drinking Water Act, enacted 1974
Bioterrorism Act-]
Public Health Security and Bioterrorism Preparedness and Response Act of 2002
enacted June 12, 2002 (PL 107-188)
ACCNSCMn
Arsenic and Clarifications to Compliance and New Source Contaminant Monitoring
promulgated Jan 22, 2001; effective Mar 23, 2001
TTHM
Total Trihalomethane Rule
promulgated Nov29,1979;
effective Nov 29,1980 for PWSs serving 75,000;
effective Nov 29,1981 for PWSs serving 10,000 to 75,000
TCR
Total Coliform Rule
RDWR
scheduled foFpromulgation
promulgated June 29,1989; effective Dec 31,1990
hSWTR
Surface Water Treatment Rule
promulgated June 29,1989; effective Dec 31,1990
NIPDWR
National Interim Primary Drinking Water Regulations
enacted between 1975 and 1976
LLCR
Lead and Copper Rule
promulgated June 7,|1991; effective Dec 7,1992
86SDWAA
Safe Drinking Water Act Amendments of 1986
enacted Jun 16,1986
ICR
Information Collection Rule
promulgated May 14,1996; effective June 18,1996
96SDWAA-
Safe Drinking Water Act Amendments of 1996
enacted Aug 6,1996
IESWTR H
Interim Enhanced Surface Water Treatment Rule
promulgated Dec 16,1998; effective Feb 16,1999
DBPR1
Stage 1 Disinfection By-Product Rule
promulgated Dec 16,1998; effective Feb 16,1999
GWR
Grouncl Water Rule
scheduled for promulgation
DBPR2
Figure 1-4. Evolution of Federal Drinking
Water Regulations - Timeline.
Stage 2 Disinfection Byproduct Rule
scheduled for promulgation
LTIESWTfiJ
Long Term 1 Enhanced Surface Water Treatment Rule
promulgated Jan 14, 2002; effective Feb 13, 2002
LT2ESWTR -
Long Term 2 Enhanced Surface Water Treatment Rule
scheduled for promulgation
Meeting and balancing the requirements of the
various regulations can provide a significant chal-
lenge to water utilities. In some cases, regulations
provide guidance or requirements that could result in
contradictory actions. For example, the SWTR
requires the use of chlorine or some other disinfectant.
However, chlorine or other disinfectants interact with
NOM in treated water to form DBFs. Similarly, raising
the pH of treated water will assist in controlling
corrosion but may increase the formation of TTHMs.
Various analytical tools, such as water quality models,
can provide the utility with information and an
understanding that helps the utility in balancing the
contradictory requirements of some regulations.
formation, may result in the sporadic occurrence of
coliforms ("indicator" organisms associated with
bacteriologically polluted water). Models can be
used to define mixing zones where blending water
from two or more sources results in water quality
problems. Specifically, water quality modeling
tools may assist utilities in complying with the
TCR, SWTR, IESWTR, LT1ESWTR, and LCR.
Modeling can assist in identifying parts of the
system with high TTHM formation potential
(DBPR1) and meeting the Initial Distribution
System Evaluation (IDSE) requirements of the
DBPR2 (see the IDSE Case Study in Chapter 7). In
addition, modeling techniques can assist in
tracking contamination from cross-connections and
other accidental or deliberate contamination events
such as a waterborne outbreak.
1.4 Assessment and Management
of Water Quality
Water utilities treat nearly 34 billion gallons of water
every day (EPA, 1999). Generally, surface water
systems require more treatment than groundwater
1-8
-------
A Reference Guide for Utilities
Table 1-1. Selected Rules and Regulations Dealing with Distribution Systems (Not Inclusive)
Regulation
SDWA
NIPDWR
TTHM
86SDWAA
TCR
SWTR
LCR
ICR
96SDWAA
IE SWTR
DBPR1
LT1ESWTR
Key Distribution System Requirements
Gives EPA the authority to establish national primary and secondary drinking water regulations
(MCLs and MCLGs).
The NIPDWR which was adopted at the passage of the SDWA required that representative
coliform samples be collected throughout the distribution system.
Established a standard forTTHMs as 0.1 mg/L.
Established the MCLG concept.
Regulates coliform bacteria which are used as surrogate organisms to indicate whether or not
treatment is effective and system contamination is occurring.
Requires using chlorine or some other disinfectant.
Monitoring for compliance with the LCR is based entirely on samples taken at the consumer's tap.
Provided data to support the interim and long-term enhanced SWTR, and Stage 2 DBP rule.
Has many provisions dealing with distribution systems, including the role that surface water
quality can play in influencing the quality of distributed water.
Provisions to enhance protection from pathogens, including Cryptosporidium, and intended to
prevent increases in microbial risk while large systems comply with the DBPR1.
Has lowered the standard for TTHMs from 0.1 mg/L to 0.08 mg/L. This standard applies to all
community water supplies in the U. S. and requires monitoring and compliance at selected points
in the distribution system.
Provisions to enhance protection from pathogens, including Cryptosporidium, and prevent
increases in microbial risk for systems serving less than 10,000 people while they comply with
theDBPRL
systems because they are directly exposed to the
atmosphere, runoff from rain and melting snow, and
other industrial sources of contamination. Water
utilities use a variety of treatment processes to remove
contaminants from drinking water prior to distribu-
tion. The selected treatment combination is based on
the contaminants found in the source water of that
particular system. The general techniques include:
• Coagulation/Flocculation: This process
removes dirt and other particles suspended in
the water. In this process, alum, iron salts, and/
or synthetic organic polymers are added to the
water to form sticky particles called "floe,"
which attract the suspended particles.
• Sedimentation: In this process, the flocculated
particles are gravity-settled and removed from
the water.
• Filtration: Many water treatment facilities use
filtration to remove the smaller particles from
the water. These particles include: clays and
silts, natural organic matter, precipitates from
other treatment processes in the facility, iron
and manganese, and microorganisms. Filtration
clarifies the water and enhances the
effectiveness of disinfection.
• Disinfection: Water is disinfected at the water
treatment plant (or at the entry to the
distribution system) to ensure that microbial
contaminants are inactivated. Secondary
disinfection is practiced in order to maintain a
residual in the distribution system.
Once the treated water enters the distribution system,
a number of processes may occur that can adversely
impact the water quality delivered to consumers. As
the water enters a network of buried pipes, valves,
joints, meters, and service lines, it is subject to
disruptions such as water hammer (transient pressure
shock wave), aging (at dead ends and large tanks),
corrosion, cross-connections, leaching of toxic
chemicals, intrusion of pathogens, and pipeline
breaks. Some of these events may be regular occur-
rences, such as water aging, loss of chlorine residual
in dead ends, or deposition of sedimentation in
stagnant areas. Others may be rare or unusual events.
Any of these events can cause the water quality to
deteriorate and pose a potential public health risk.
Some routine distribution system design changes and
maintenance or operational procedures that can help
to prevent or reduce the effects of such events include
the following:
• Tank Mixing: Inadequate mixing in a tank can
lead to stagnant areas containing older water
1-9
-------
A Reference Guide for Utilities
Maintaining water quality in a drinking water distribu-
tion system while assuring adequate disinfection and
reducing DBFs is a significant challenge for many
drinking water utilities. This challenge will be even
greater under the more stringent requirements of the
LT2ESWTR and the DBPR2. Utilities that use chorine
as their primary disinfectant and that have elevated
organic levels in their treated water, long detention
times, and/or warm water may have difficulty in meeting
these regulations. The Las Vegas Valley Water District is
conducting research to explore the feasibility of employ-
ing "targeted" distribution system treatment systems.
This type of targeted system (or systems) would utilize
small-scale water treatment technology to reduce the
concentration of disinfection byproducts in those areas
that might exceed the SDWA MCLs established under
the LT2ESWTR and DBPR2. These systems are in-
tended to be designed and operated in conjunction with
a water quality/hydraulic model which would be used to
predict where these decentralized treatment systems
should be located. If the treatment technology is
relatively mobile, it could be moved based on model
predictions to locations where MCL violations are
likely to occur. In addition, these types of systems
would be valuable should a security threat arise.
that has lost its disinfectant residual. Changes
in operations (e.g., exercising the tank) or
modifications to inlet-outlet configurations can
improve mixing.
Re-chlorination: Some parts of a distribution
system may experience long travel times from
the treatment plant resulting in loss of chlorine
residual. Installation of booster chlorination
facilities at these locations can sometimes be an
effective means of insuring an adequate residual
in these areas.
Conventional Flushing: This procedure
generally involves opening hydrants in an area
until the water visibly runs clear. The object of
this action would be to quickly remove
contaminated water; however, it would not
likely be effective in removal of contaminants
that become attached to the pipe surfaces.
Flushing only provides a short-term remedy.
Unidirectional Flushing: This procedure
involves the closure of valves and opening of
hydrants to concentrate the flow in a limited
number of pipes. Flow velocities are
maximized so that shear velocity near the pipe
wall is maximized. It is intended to be done in
a progressive fashion, proceeding outward from
the source of water in the system so that
flushing water is drawn from previously flushed
Federal and state drinking water regulations are designed
to provide a water supply to consumers that meets
minimum health-based requirements. However, water
utilities may choose to implement programs that go
beyond current federal, state, and local regulatory
requirements to increase the water quality and reduce the
potential for contamination in water systems. There are
several methods and guidelines that have been designed
to assist utilities in providing water of a quality that
exceeds the minimum requirements. These methods
include: Hazard Analysis Critical Control Point
(HACCP), source water optimization, and distribution
system water quality optimization plans (DSOP).
DSOP is one example of a framework for evaluating and
improving programs that affect distribution system water
quality (Friedman et al., 2005). Aspects of the DSOP
include evaluation of conditions within the distribution
system, creation of improved documentation, and
enhancement of communication between the various
utility functions that impact water quality in the distri-
bution system. DSOPs address both regulatory/compli-
ance issues and customer issues related to aesthetic
properties of drinking water. The D SOP approach was
piloted at three water utilities and a general template was
developed that can be used by small, medium, and large
utilities. The following ten steps are identified as part of
the development of a DSOP:
1. Formation of a committee to discuss distribution
system issues of interest/concern and to guide the
process of DSOP development.
2. Identification of water quality and operating goals.
3. Completion of a distribution system audit.
4. Comparison of audit results to industry best
management practices.
5. Development of a list of utility needs for
optimizing distribution system water quality.
6. Prioritization of DSOP elements based on relative
contribution towards improving water quality and
precluding water quality degradation or
contamination.
7. For each priority DSOP element, compilation of
applicable standard operating procedures (SOPs)
and ongoing programs that provide information
related to the condition of the distribution system
and water quality.
8. Development and implementation of priority
programs.
9. Periodic review of programs and goals developed
as part of the DSOP.
10. Development of revised SOPs that describe the
optimized approach.
DSOP and other aforementioned methodologies are still
in their early stages of application in the water supply
industry and will require further evaluation to determine
their effectiveness in meeting the goals to improve water
quality in drinking water systems.
1-10
-------
A Reference Guide for Utilities
reaches. No special equipment is required;
however, some planning time is required to
determine the flushing zones, the valves and
hydrants to be operated, and the duration of the
flushing exercise for each zone.
Valve Exercising Program: A routine program to
exercise isolation valves can have several
positive effects. These include identifying (and
repairing) malfunctioning valves and
identifying valves that are in an inappropriate
setting (e.g., closed valves that are expected to
be open).
Cross-Connection Control Program: An
inspection program intended to ensure no
interconnection(s) between the drinking water
and wastewater systems in homes and buildings.
Examples of routine maintenance and operation
procedures for pipe cleaning include the following
(AwwaRF, 2004):
• Air Scouring, Swabbing and Abrasive pigging: Air
scouring, swabbing, and abrasive pigging are
progressively more aggressive cleanup techniques
that involve more specialized equipment and skills.
A few water utilities have implemented these
methods using their own staff; typically, these
methods are contracted to specialty firms.
Implementation of these methods would require
installation of new pipeline appurtenances (e.g., pig
launching and receiving stations; pigging is not
recommended for cast iron pipes).
• Chemical/Mechanical Cleaning and Lining:
Chemical cleaning involves the recirculation in an
isolated pipe section of proprietary acids and
surfactants to remove scale and deposits, while
mechanical cleaning is accomplished by dragged
scrapers. These techniques are typically applied in
the rehabilitation of older unlined cast iron pipe
which, over time, have become scaled and
tuberculated. These cleaning operations are
typically followed by an in-situ application of a
thin cement mortar or epoxy lining to ensure lasting
protection.
If the symptoms persist after the application of these
techniques, the pipes are usually replaced.
1.5 Advanced Tools for Water
Quality Management
Recent advancements in computation and instru-
mentation technologies have led to the availability
of advanced tools that are already beginning to
improve a utility's ability to effectively manage
water quality in distribution systems. These
computational advancements have led to the
development of software models that can simulate
the behavior of distribution system networks. Water
distribution system models (such as EPANET) have
become widely accepted both within the water
utility industry and the general research arena for
simulating both hydraulic and water quality
behavior in water distribution systems. The
advancements in instrumentation and Supervisory
Control and Data Acquisition (SCADA) systems
now enable the utilities to monitor and control
various water quality parameters from a remote
location in real-time within a distribution system
network. Furthermore, recent advances in Geo-
graphic Information Systems (GIS) technology have
led to the integration of remote monitoring network
models with GIS layers. This combination provides
utilities a visual tool to efficiently manage both
water quality and distribution system assets such as
pipes, pumps, and valves.
1.6 Report Organization
Various chapters of this reference guide will
describe modeling and monitoring tools for
effectively managing water quality in drinking
water distribution systems. Examples and protocols
for effectively applying water quality models for
understanding and resolving water quality issues in
networks will be presented. Another important
aspect of effectively applying water quality models
is to ensure that they are properly and periodically
calibrated. Tracer tests are one of the most effec-
tive techniques for calibrating a water quality
model. Modeling techniques, when combined with
advanced monitoring and geospatial technologies.
can play a vital role in managing water quality in
distribution systems. Chapter 2 provides an
overview on modeling of distribution systems for
water quality. Chapter 3 describes techniques for
conducting tracer studies in distribution systems.
Chapter 4 presents data analysis techniques for
effectively calibrating a distribution system model
using tracer or other field data. Chapter 5 provides
an overview of monitoring techniques and tech-
nologies available for monitoring water quality.
Chapter 6 introduces geospatial technology and its
relation to water distribution systems. Finally.
Chapter 7 is a compilation of real-world applica-
tions of water quality modeling and monitoring for
planning, analysis and simulation of historical
events.
1.7 Summary
Distribution system infrastructure is a major asset of
most water utilities. It serves many important
functions in a community, such as promoting eco-
1-11
-------
A Reference Guide for Utilities
The information presented in this reference guide is
intended for a general technical audience. The
various chapters provide an overview of the state-of-
art techniques for managing water quality in distribu-
tion systems. For a more comprehensive case-specific
solution, the reader should refer to text books in
specific subject areas and/or consult with water
quality professionals. The following is a brief listing
of recommended books (listed in alphabetical order
by title):
1. Advanced Water Distribution Modeling and
Management. T.M. Walski, D.V Chase, D.A.
Savic, W. Grayman, S. Beckwith, andE. Koelle.
Haestad Press, Waterbury, CT. 2003.
2. Comprehensive Water Distribution Systems
Analysis Handbook. P.B. Boulos, K.E. Lansey,
andB.W. Karney. MWHSOFT, Inc., Pasadena, CA.
2004.
3. Computer Modeling of Water Distribution
Systems (M32), AWWA. 2004.
4. GIS Applications for Water, Wastewater, and
Stormwater Systems. U. Shamsi. CRC Press. 2005.
5. Hydraulics of Pipeline Systems. B.E. Larock, R.W
Jeppson, G.Z. Walters. CRC Press. 1999.
6. Microbial Quality of Water Supply in Distribution
Systems. Edwin E. Geldreich. CRC Press. 1996
7. Modeling, Analysis and Design of Water
Distribution Systems. L. Cesario. AWWA. 1995.
8. Modeling Water Quality in Drinking Water
Distribution Systems. R.M. Clark and WM.
Grayman. AWWA. 1998.
9. Online Monitoring for Drinking Water Utilities.
Edited by E. Hargesheimer, O. Conio, and J.
Popovicova. AwwaRF - CRS ProAqua. 2002.
10. Safe Drinking Water: Lessons from Recent
Outbreaks in Affluent Nations. S.E. Hrudey and
E.J. Hrudey, IWA Publishing. 2004.
11. Water Distribution Systems Handbook. Edited by
L.W. Mays, McGraw Hill. 2000.
12. Water Supply Systems Security. Edited by L.W
Mays, McGraw Hill. 2004.
nomic growth, supporting public safety, and protect-
ing public health. In order for a community to grow
and prosper, it must have the physical infrastructure to
provide basic services such as water supply. In
addition to the economic implications of adequate
water supply, water systems play a critical role in
supporting public safety through the provision of fire
protection capacity. Frequently, insurance rates in a
community are tied to the fire protection capability of
the water system. Water systems play a key role in
protecting a community's public health by providing
safe drinking water to water consumers.
1-12
-------
References
A Reference Guide for Utilities
American Society of Civil Engineers (ASCE).
"Report Card for America's Infrastructure, Drinking
Water." ASCE, Reston, VA. 2005.
American Water Works Association (AWWA). AWWA
Manual M31, Distribution System Requirements for
Fire Protection, AWWA, Denver, CO. 1998.
AWWA. Water Quality & Treatment-A Handbook of
Community Water Supplies. Edited by R.D.
Letterman, McGraw-Hill, NY. 1999.
American Water Works Association Research Founda-
tion (AwwaRF). Standard Operating Procedures for
Decontamination of Water Infrastructure. March,
2004.
AwwaRF. Impacts of Fire Flow on Distribution
System Water Quality, Design, and Operation. 2002.
Camper, A.K., K. Brastrup, A. Sandvig, J. Clement, C.
Spencer, and A. J. Capuzzi. "Impact of distribution
system materials on bacterial regrowth." Journal of
AWWA, 95(7):107-121. 2003.
Clark, R.M., WM. Grayman, and R.M. Males.
"Contaminant Propagation in Distribution Systems,"
Journal of Environmental Engineering, ASCE,
114:(2). 1988.
Clark, R.M., WM. Grayman, and J.A. Goodrich.
"Water Quality Modeling: Its Regulatory Implica-
tions." Proceedings, AwwaRF/EPA Conference on
Water Quality Modeling in Distribution Systems,
Cincinnati, OH. 1991a.
Clark, R.M., D.J. Ehreth, and J.J. Convery. "Water
Legislation in the US: An Overview of the Safe
Drinking Water Act," Toxicology and Industrial
Health, 7(516):43-52. 1991b.
Clark, R.M., and W.A. Feige. "Meeting the Require-
ments of the Safe Drinking Water Act," in Strategies
and Technologies for Meeting the Requirements of
the SDWA. Edited by Robert M. Clark and R. Scott
Summers, Technomics, Publishing Company, Inc.,
Lancaster, PA. 1993.
Clark, R.M., W.M. Grayman, R.M. Males, and A.F.
Hess, 1993a. "Modeling Contaminant Propagation in
Drinking Water Distribution Systems," Journal of
Environmental Engineering, ASCE, 119(2):349-364.
1993.
Clark, R.M., J.A. Goodrich, andL.J. Wymer. "Effect of
the Distribution System on Drinking Water Quality,"
Journal of Water Supply Research and Technology -
Aqua, 42(l):30-38. 1993b.
Clark, R.M., E.E. Geldreich, K.R. Fox, E.W Rice, C.H.
Johnson, J.A. Goodrich, J.A. Barnick, andF.
Abdesaken. "Tracking a Salmonella serovar
Typhimurium Outbreak in Gideon, Missouri: Role of
Contaminant Propagation Modeling," Journal of
Water Supply Research and Technology - Aqua,
45(4):171-183. 1996.
Clark, R.M., and W.M. Grayman. Modeling Water
Quality in Drinking Water Distribution Systems.
AWWA, Denver, CO. 1998.
Clark, R.M., W.M. Grayman, S.G. Buchberger, Y. Lee,
and D.J. Hartman. "Drinking Water Distribution
Systems: An Overview," in Water Supply Systems
Security. Edited by Larry W. Mays, McGraw-Hill, NY.
pp 4.1-4.49. 2004
Clark, R.M., and D. L. Tippen. "Water Supply" in
Standard Handbook of Environmental Engineering.
Edited by Robert A. Corbitt, McGraw-Hill, NY. pp
5.173-5.220. 1990.
Connell, G.F. "European Water Disinfection Practices
Parallel U.S. Treatment Methods," Drinking Water
and Health Quarterly. August 1, 1998. http://
www. clo2. com/reading/waternews/european. html
EPA. The Clean Water and Drinking Water Infra-
structure Gap Analysis. EPA - Office of Water,
Washington DC. September 2002.
EPA. 25 Years of the Safe Drinking Water Act:
History & Trends. EPA816-R-99-007. December
1999.
Fair, G.M., and J.C. Geyer. "Water Supply and Waste-
Water Disposal," John Wiley and Sons, Inc., NY. 1971.
Friedman, M., G. Kirmeyer, G. Pierson, S. Harrison, K.
Martel, A. Sandvig, and A. Hanson. Development of
Distribution System Water Quality Optimization
Plans. AwwaRF/AWWA. Denver, CO. 2005.
Geldreich, E.E., H.D. Nash, D.J. Reasoner, andR.H.
Taylor. "The Necessity of Controlling Bacterial
Populations in Potable Water: Community Water
Supply," Journal of AWWA, 64:596-602. 1972.
1-13
-------
A Reference Guide for Utilities
Greater Cincinnati Water Works. "Water Treatment,"
http ://www. cincinnati-oh. gov/water/pages/-3 283-7.
2005.
Grigg, NS. Assessment and Renewal of Water
Distribution Systems. AwwaRF Report # 91025F,
June 2005.
Gullick, R.W., W.M. Grayman, R.A. Deininger, and
R.M. Males. "Design of Early Warning Monitoring
Systems for Source Waters," Journal ofAWWA,
95(ll):58-72. 2003.
Insurance Services Office (ISO). Fire Suppression
Rating Schedule. Insurance Services Office, NY.
2003.
International Life Sciences Institute. Early Warning
Monitoring to Detect Hazardous Events in Water
Supplies. Edited by T.M. Brosnan, Washington, D.C.
1999.
Kirmeyer, G.J., W Richards, andC. Dery-Smith. An
Assessment of Water Distribution Systems and
Associated Needs. AwwaRF. 1994.
LeChevallier, M.W, T.M. Babcock, andR.G. Lee.
"Examination and Characterization of Distribution
System Biofilms," Applied and Environmental
Microbiology, 53:2714-2724. 1987.
Maul, A., A.H. El-Shaarawi, and J.C. Block. "Het-
erotrophic Bacteria in Water Distribution Systems -1.
Spatial and Temporal Variation." The Science of the
Total Environment, 44:201-214. 1985a.
Maul, A., A.H. El-Shaarawi, and J.C. Block. "Het-
erotrophic Bacteria in Water Distribution Systems - II.
Sampling Design for Monitoring." The Science of the
Total Environment, 44:215-224. 1985b.
Montana State University (MSU) Center for Biofilm
Engineering. Image provided by Pat Dirckx. 2005.
National Fire Protection Association (NFPA). Fire
Protection Handbook, 19th edition. Edited by A.E.
Cote, National Fire Protection Association, Quincy,
MA. 2003.
NRC. Public Water Supply Distribution Systems:
Assessing and Reducing Risks - First Report. The
National Academies Press, Washington, DC. 2005.
Okun, D. "Distributing reclaimed water through dual
systems," Journal ofAWWA, 89 (11), pp. 52-64. 1996.
Owens, J. A Review of Federal Drinking Water
Regulations in the U.S., in Controlling Disinfection
By-Products and Microbial Contaminants in Drinking
Water. Edited by Robert M. Clark and Brenda Boutin,
EPA-600-R-01-110. pp 2-1 to 2-14. December 2001.
FDD 63. Critical Infrastructure Protection. The
White House, Washington D.C. May 22, 1998.
PL107-188. Public Health Security and Bioterrorism
Preparedness and Response Act of 2002.
Safe Drinking Water Act (SDWA), 1974, Public Law
93-523.
Walski, T. M, D.V. Chase, D.A. Savic, W.M. Grayman,
S. Beckwith, and E. Koelle. Advanced Water Distri-
bution Modeling and Management. Haestad Press,
Waterbury, CT pp 1-4. 2003.
Water Research Centre. "Deterioration of Bacterio-
logical Quality of Water During Distribution," Notes
on Water Research No. 6. 1976.
Zhang, W., and F.A. DiGiano. "Comparison of
Bacterial Regrowth in Distribution Systems Using
Free Chlorine and Chloramine: A Statistical Study of
Causative Factors." Water Research, 36:6:1469-1482.
2002.
1-14
-------
A Reference Guide for Utilities
Chapter 2
Modeling Water Quality in Drinking Water
Distribution Systems
This chapter covers the use of models to simulate the flow and water quality conditions in a distribution
system network. Models are mathematical or physical approximations of a real-world system and can be
used to study the behavior of actual system(s). A variety of computer software modeling tools are now
available to perform these simulations. These tools are now commonly used by trained engineers and
scientists to study and improve water distribution system network design and operation.
Water distribution system models have become
widely accepted within the water utility industry as a
mechanism for simulating the hydraulic and water
quality behavior in water distribution system net-
works. Current water distribution modeling software
is powerful, sophisticated and user-friendly. Many
software packages are integrated with GIS and
Computer Aided Design (CAD) technology in order to
facilitate model construction and storage and display
of model results. Early network models simulated
only steady-state hydraulic behavior. In the 1970s,
modeling capability was expanded to include
Extended Period Simulation (EPS) models that could
accommodate time-varying demand and operations.
Subsequently, in the early 1980s, investigators began
introducing the concept of water quality modeling.
Most water distribution system modeling software
packages now routinely incorporate water quality
simulation capability. More recently, transient
models for simulating water hammer (a transient
phenomenon) and tank mixing/aging models have
either been incorporated into or integrated with water
distribution system models. Algorithms have been
developed that enable users to optimize water system
design and operation, assist in model calibration, and
perform probabilistic analyses. Each of these model
types are briefly described later in this chapter.
Water distribution system models are more commonly
being used to replicate the behavior of a real or
proposed system for a variety of purposes including:
capital investment decisions, development of master
plans, estimation of fire protection capacity, design of
new systems and extension or rehabilitation of
existing systems, energy management, water quality
studies, various event simulations and analysis,
optimal placement of sensors, and daily operations.
The costs associated with constructing and maintain-
ing a distribution system model may be more easily
justified if it is used for a variety of applications by a
water utility (Grayman, 2000).
2.1 Distribution System Network
Hydraulic Modeling
The network hydraulic model provides the foundation
for modeling water quality in distribution systems.
This subsection provides a brief history of hydraulic
modeling, an overview of theoretical concepts, basic
model inputs, and general criteria for selection and
application.
2.1.1 History of Hydraulic Modeling
Hardy Cross first proposed the use of mathematical
methods for calculating flows in complex networks
(Cross, 1936). This manual, iterative procedure was
used throughout the water industry for almost 40
years. With the advent of computers and computer-
based modeling, improved solution methods were
developed for utilizing the Hardy Cross methodology.
The improved implementations of this method were in
widespread use by the 1980s (Wood, 1980a).
Also, in the early 1980s, the concept of modeling
water quality in distribution system networks was
developed based on steady-state formulations (Clark
et al., 1986). By the mid-1980s, water quality models
were developed that incorporated the dynamic
behavior of water networks (Grayman et al., 1988).
The usability of these models was greatly improved in
the 1990s with the introduction of the public domain
EPANET model (Rossman, 2000) and other Windows-
based commercial water distribution system models.
Initially, hydraulic models simulated flow and
pressures in a distribution system under steady-state
conditions where all demands and operations re-
mained constant. Since system demands (and
consequently the flows in the water distribution
network) vary over the course of a day, EPS models
were developed to simulate distribution system
behavior under time-varying demand and operational
conditions. These models have now become ubiqui-
tous within the water industry and are an integral part
of most water system design, master planning, and fire
flow analyses.
2-1
-------
A Reference Guide for Utilities
2.1.2 Overview of Theoretical Concepts
The theory and application of hydraulic models is
thoroughly explained in many widely available
references (Walski et al., 2003; American Water Works
Association, 2004; Larock et al., 2000). Essentially,
three basic relations are used to calculate fluid flow in
a pipe network. These relationships are:
• Conservation of Mass: This principle requires
that the sum of the mass flows in all pipes
entering a junction must equal the sum of all
mass flows leaving the junction. Because water
is essentially an incompressible fluid,
conservation of mass is equivalent to
conservation of volume.
• Conservation of Energy: There are three types
of energy in a hydraulic system: kinetic energy
associated with the movement of the fluid,
potential energy associated with the elevation,
and pressure energy. In water distribution
networks, energy is referred to as "head" and
energy losses (or headlosses) within a network
are associated primarily with friction along pipe
walls and turbulence.
• Pipe Friction Headloss: A key factor in
evaluating the flow through pipe networks is
the ability to calculate friction headloss
(Jeppson, 1976). Three empirical equations
commonly used are the Darcy-Weisbach, the
Hazen-Williams, and the Manning equations.
All three equations relate head or friction loss in
pipes to the velocity, length of pipe, pipe
diameter, and pipe roughness. A fundamental
relationship that is important for hydraulic
analysis is the Reynolds number, which is a
function of the kinematic viscosity of water
(resistance to flow), velocity, and pipe diameter.
The most widely used headloss equation in the
U.S. is the Hazen-Williams equation. Though
the Darcy Weisbach equation is generally
considered to be theoretically more rigorous,
the differences between the use of these two
equations is typically insignificant under most
circumstances.
A distribution system is represented in a hydraulic
model as a series of links and nodes. Links represent
pipes whereas nodes represent junctions, sources,
Hydraulic models represent the basic underlying
equations (conservation of mass and conservation of
energy) as a series of linear and non-linear equations.
Because of the non-linearity, iterative solution methods
are commonly used to numerically solve the set of
equations. The most common numerical method
utilized is the Newton-Raphson method.
Figure 2- 7. Simple Link-Node Representation of a
Water Distribution System.
tanks, and reservoirs. Valves and pumps are repre-
sented as either nodes or links depending on the
specific software package. Figure 2-1 illustrates a
simple link-node representation of a water distribu-
tion system.
As mentioned previously, there are two types of
analyses that may be conducted on drinking water
distribution systems: steady-state and EPS. In a
steady-state analysis, all demands and operations are
treated as constant over time and a single solution is
generated. In the EPS mode, variations in demand,
tank water levels, and other operational conditions
are simulated by a series of steady-state analyses that
are linked together. Each steady-state solution in the
EPS mode involves a separate solution of the set of
non-linear equations. EPS is used as the basis for
Conservation of Mass: The conservation of mass
principle for hydraulic analysis requires that the sum of
the mass flow in all pipes entering a junction must
equal the sum of all mass flows leaving the junction.
In EPS, if storage is involved, a term for describing the
accumulation of water at those nodes is included.
Mathematically, the principle can be represented as
follows:
2! (2 -{/,)- — = 0 (Equation 2-1)
1-1 &
where
Qt = inflow to node in i-th pipe in ftVsec (m3 /sec)
Uj = water used or leaving at the i-th node in ft3/sec (m3 /sec)
— = change in storage in ft3/sec (m3 / sec)
dt
2-2
-------
A Reference Guide for Utilities
water quality modeling. Though the EPS solution
does introduce some approximations and ignores the
transient phenomena resulting from sudden changes
(e.g., a pump being turned on), these more refined
assumptions are generally not considered significant
for most distribution system studies.
Conservation of Energy: The conservation of energy
principle requires that the difference in energy
between two points in a network must be the same
regardless of flow path. For hydraulic analysis, this
principle can be represented in terms of head as
follows:
5. + ^. +£/2p =Z2 +^.+^-
7 2g 7 2g
\ +2>M (Equation 2-2)
where
Zland2 = elevation at points 1 and 2, respectively, inft (m)
Pland2 = pressure at points 1 and 2, respectively, in lb/ft2 (N/m2)
f = fluid (water) specific weight, in lb/ft3 (N/m3)
Kland2 = velocity at points 1 and 2, respectively, in ft/s (m/s)
g = acceleration due to gravity, in ft/sec2 (m/sec2)
hp = pumping head gain, inft (m)
hL = head loss in pipes, inft (m)
hM = head loss due to minor losses, in ft (m)
Pipe-friction headloss: The equation most commonly
used in modeling software for computation of pipe-
friction headloss is the Hazen-Williams equation
represented as follows:
1.85,^4.87 Q
(Equation 2-3)
where
hL = head loss due to friction, inft (m)
C( — Unit conversion factor
(4.73 in British units; 10.7 in Metric units)
D = pipe diameter, inft (m)
L — length of pipe, in ft (m)
Q = pipe flow rate, in ftVsec (mVsec)
C — Hazen-Williams coefficient (dimensionless)
2.1.3 Basic Hydraulic Model Input
Characterization
Building a network model, particularly if a large
number of pipes are involved, is a complex process.
The following categories of information are needed to
construct a hydraulic model:
• Characteristics of the pipe network components
(pipes, pumps, tanks, valves).
• Water use (demands) assigned to nodes
(temporal variations required in EPS).
• Topographic information (elevations assigned
to nodes).
• Control information that describes how the
system is operated (e.g., mode of pump
operation).
• Solution parameters (e.g., time steps, tolerances
as required by the solution techniques).
Commonly used methods for these inputs are briefly
described in the following subsections.
2.1.3.1 Pipe Network Inputs
Construction of the pipe network and its characteris-
tics may be done manually or through use of existing
spatial databases stored in GIS or CAD packages.
Most commonly, GIS or CAD packages are used in
this process and are described in more detail in
Chapter 6. The initial step in constructing a network
model is to identify pipes to be included in the
model. Nodes are usually placed at pipe junctions, or
at major facilities (tanks, pumps, control valves), or
where pipe characteristics change in diameter, "C"-
value (roughness), or material of construction. Nodes
may also be placed at locations of known pressure or
at sampling locations or at locations where water is
used (demand nodes). The required pipe network
component information includes the following:
• pipes (length, diameter, roughness factor),
• pumps (pump curve),
• valves (settings), and
• tanks (cross section information, minimum and
maximum water levels).
2.1.3.2 Water Demand Inputs
Water consumption or water demand is the driving
force behind the operation of a water distribution
system. Any location at which water leaves the
system can be characterized as a demand on the
system. The water demands are aggregated and
assigned to nodes, which represents an obvious
simplification of real-world situations in which
individual house taps are distributed along a pipe
rather than at junction nodes. It is important to be
able to determine the amount of water being used,
where it is being used, and how this usage varies with
time (Walski et al., 2003). Demand may be estimated
by a count of structures of different types using a
representative consumption per structure, meter
readings and the assignment of each meter to a node,
and to general land use. A universal adjustment factor
should be used to account for losses and other
unaccounted water usage so that total usage in the
2-3
-------
A Reference Guide for Utilities
Early software packages limited the number of pipes that
could be included due to computer storage restrictions.
This led to the concept of "skeletonizing" a network or
including only those pipes that were considered to be
the most important. The degree of skeletonization that is
acceptable should depend upon the ultimate use of the
model. For example, master plans and energy studies
might be based on the use of skeletonized networks.
Other applications, such as water quality modeling and
designing flushing programs, require a model that
includes more pipes. Though there is no national
standard for skeletonization, the EPA draft guidance
issued for modeling to support the IDSE under DBPR2
suggests inclusion of (EPA, 2003):
• At least 50 percent of total pipe length in the
distribution system.
• At least 75 percent of the pipe volume in the
distribution system.
• All 12-inch diameter and larger pipes.
• All 8-inch and larger pipes that connect pressure
zones, influence zones from different sources,
storage facilities, major demand areas, pumps, and
control valves, or are known or expected to be
significant conveyors of water.
• All 6-inch and larger pipes that connect remote
areas of a distribution system to the main portion
of the system.
• All storage facilities with controls or settings
applied to govern the open/closed status of the
facility that reflect standard operations.
• All active pump stations with realistic controls or
settings applied to govern their on/off status that
reflect standard operations.
• All active control valves or other system features
that could significantly affect the flow of water
through the distribution system (e.g.,
interconnections with other systems, valving
between pressure zones).
A case study presented in Section 7.3.1 illustrates the
use of models in support of IDSE.
Most modern software packages support an unlimited
number of pipes; however, skeletonization is still
frequently used in order to reduce the modeling effort. A
minimal skeletonization should include all pipes and
features of major concern.
model corresponds to total production.
In order to use a model in the EPS mode, information
on temporal variations in water usage over the period
being modeled are required. Spatially different
temporal patterns can be applied to the individual
network nodes. The best available information
should be used for developing temporal patterns in
order to make EPS most effective. For example, some
users may have continuous water metering data, while
others may use literature values as a first approxima-
tion for estimating residential temporal patterns.
Temporal patterns also vary with climate. For ex-
ample, lawn watering in summer months will cause a
spike in usage of water during that time period. In
some cases, information from SCADA systems can be
used to estimate system-wide temporal patterns.
A typical hierarchy for assigning demands includes
the following:
• Baseline Demands: Baseline demands usually
correspond to consumer demands and
unaccounted-for-water associated with average
day conditions. This information is often
acquired from a water utility's existing records,
such as customer meter and billing records.
Although the spatial assignment of these
demands is extremely important and should
include the assignment of customer classes such
as industrial, residential, and commercial use,
actual metering data should be used when
available.
• Seasonal Variation: Water use typically varies
over the course of the year with higher demands
occurring in warmer months. When developing
a steady-state model, the baseline (average day)
demand can be modified by multipliers in order
to reflect other conditions such as maximum
day demand, peak-hour demand, and minimum
day demand.
• Fire Demands: Water provided for fire services
can be the most important consideration in
developing design standards for water systems.
Typically, a system is modeled corresponding to
maximum-use conditions, with needed fire-flow
added to a single node at a time. It is not
uncommon for a requirement that multiple
hydrants be flowing simultaneously.
• Diurnal Variation: All water systems are
unsteady due to continuously varying demands.
It is important to account for these variations in
order to achieve an adequate hydraulic model.
Diurnal varying demand curves should be
developed for each major consumer class or
geographic zones within a service area. For
example, diurnal demand curves might be
developed for industrial establishments,
commercial establishments, and residences.
Large users such as manufacturing facilities
may have unique usage patterns.
2-4
-------
A Reference Guide for Utilities
Future water use: For design and planning purposes, a
water system must be examined under future conditions.
In situations where a system is largely currently built
out, future demands may be estimated by developing
global or regional multipliers that are applied to current
demands. However, in new or developing areas, existing
water use does not provide a useful basis for estimating
future demands. Alternative approaches use popula-
tion-based projections, socioeconomic modeling, and
land-use methods (Johnson and Loux, 2004).
In estimating future demands for use in a network
model, the most appropriate method is generally the
land-use method. The land-use method is based on
mapping land uses and then applying a water-use factor
to each land-use category. When applied to existing
situations or in historical reconstruction of water
systems, aerial photographs are most commonly used as
the base map for identifying land-use categories. For
development of future demands, land-use maps can be
obtained from planners. The land-use methodology is
depicted in Figure 2-2.
r
Determine
land use
categories
Develop future
polygons within
sphere of
influence
I
y unit
s to land
Dlygons
->
1
Develop
existing
polygons
Identify
unique
water users
i
Phase
future
polygons
J
Water
Demands
Figure 2-2. A Flow Chart for Estimating Future Water
Demand Based on Land-Use Methodology.
Land-use unit demands or water-use factors are typically
developed in units of gallons per day (GPD) per acre
from local historical consumption data or from available
regional information. GIS technology is frequently
used as a means of developing and manipulating the
land-use polygons and assigning the calculated de-
mands to the model nodes.
2.1.3.3 Topographical Inputs
Hydraulic models use elevation data to convert heads
to pressure. Actual pipe elevations should be used to
establish the correct hydraulic gradeline. Elevations
are assigned to each node in a network where pressure
information is required. Various techniques are used
to determine elevation information including the
following:
• Topographical maps: Paper topographical maps
produced by the United States Geological
Survey (USGS) or other local agencies may be
used to manually interpolate elevations for
nodes. The relative accuracy depends upon the
degree of topography in the area, the contour
elevations on the map, and the manual takeoff
methods used.
• Digital elevation models (DEM): USGS and
other agencies produce digital files containing
topographical information. When used with
various software tools, elevation information
can be directly interpolated and assigned to
nodes based on the coordinates of the nodes.
The accuracy of this process depends upon the
degree of detail in the DEM.
• Global Positioning Systems (GPS) or other field
survey methods: Standard field surveying
techniques or modern surveying methods using
a GPS satellite can be used to measure
elevations at nodes. The modern GPS units can
calculate elevation by using four or more
satellites. However, elevation is the most
difficult calculation for a GPS unit, and
depending upon the location surveyed, it may
be prone to significant error.
2.1.3.4 Model Control Inputs
In order to apply an EPS model, it is necessary to
define a set of rules that tells the model how the water
system operates. This may be as simple as specifying
that a particular pump operates from 7:00 AM to
10:00 AM each day. Alternatively, it may be a set of
complex "logical controls" in which operations such
as pump off/on, pump speed, or valve status are
controlled using Boolean operators (including if-
then-else logic) for factors such as tank water levels,
node pressures, system demand, and time of day
(Grayman and Rossman, 1994). For water systems
that operate automatically based on a set of rules,
determination of these rules are quite straightforward.
For manual systems, the rules must be determined by
interviews with system operators.
2.1.3.5 Extended Period Simulation (EPS)
Solution Parameters
Solution techniques used to iteratively solve the set
of non-linear equations typically have various global
parameters that control the solution technique. These
parameters may be time-step lengths for EPS runs or
tolerance factors that tell the model when a solution is
considered to have converged. The user must specify
2-5
-------
A Reference Guide for Utilities
the values for the solution parameters, or (as is
frequently done) accept the default values that are
built into the software products. The specific solution
parameters vary between solution techniques and
specific software products.
2.1.4 General Criteria for Model Selection and
Application
The initial step in modeling is to define the basic
scope and needs of the modeling process and to select
an appropriate software package that will satisfy both
the specific needs of the current project and likely
future needs. Factors that may enter into the selection
of a software package include:
• technical features,
• training/support and manuals,
• user interface,
• integration with other software (such as GIS,
CAD),
• compatibility with EPANET,
• cost, and
• response from existing users.
A summary of major available hydraulic-water quality
modeling software is provided in Section 2.3.2. Once
a suitable model has been selected, the following
steps should be followed in applying network models
(Clark and Grayman, 1998):
• Develop the basic network model.
• Calibrate and validate the model.
• Establish clear objectives and apply the model
in a manner to meet the objectives.
• Analyze and display the results.
2.1.4.1 Developing a Basic Network Model
The basic network model inputs should be first
characterized using the techniques described in
Section 2.1.3. The model should be developed based
on accurate, up-to-date information. Information
should be entered carefully and checked frequently.
Following the entry of the data, an initial run of the
model should be made to check for reasonableness.
2.1.4.2 Model Calibration and Validation
Calibration is an integral aspect of the art of modeling
water distribution systems. Model calibration is the
process of adjusting model input data (or, in some
cases, model structure) so that the simulated hydraulic
and water quality output sufficiently mirrors observed
field data. Depending on the degree of accuracy
desired, calibration can be difficult, costly, and time-
consuming. The extent and difficulty of calibration
are minimized by developing an accurate set of basic
inputs that provide a good representation of the real
network and its components.
A traditional technique for calibration is the use of
"fire-flow" tests. In a fire-flow test, the system is
stressed by opening hydrants to increase flows in
small parts of the system. This results in increased
headloss in pipes in the vicinity of the test. Pressures
and flow are then measured in the field. Model
parameters, such as roughness factors (C), demands,
and valve positions, are adjusted so that the model
adequately reflects the field data. Another common
calibration technique is to measure predicted tank/
reservoir levels derived from computer simulations
against actual tank levels during a given period of
record. For example, using water level, pressure, or
flow data from SCADA systems or from on-line
pressure and tank-level recorders, model parameters
(such as roughness, water demands, and pump
controls) can be adjusted in the simulation model
until the model results match the actual tank level
and other continuous information for the defined
criteria. The resulting optimal parameter values
should be checked to ensure that the values are
realistic. Sophisticated commercial hydraulic models,
such as those listed in Section 2.4, may also incorpo-
rate optimization components that aid the user in
selecting system parameters resulting in the best
match between observed system performance and
model results (Walski, 2003).
Model validation is the step that follows calibration
and uses an independent field data set to verify that
the model is well calibrated. In the validation step,
the calibrated model is run under conditions differing
from those used for calibration and the results
compared to field data. If the model results closely
approximate the field results (visually) for an appro-
priate time period, the calibrated model is considered
to be validated. Significant deviations indicate that
further calibration is required. A variety of calibra-
tion and validation techniques suitable to both large
and small water utilities are discussed in Chapter 4 of
this document.
Another rigorous methodology for calibration and valida-
tion is the use of tracers. Concentrations of naturally
occurring materials or added chemical tracers may be
measured in the field and the results used to calibrate
hydraulic and water quality models. This methodology is
further described in Chapter 3 of this document.
2.1.4.3 Establishing Objectives and Model
Application
Prior to applying the model, the specific modeling
objectives should be clearly established. The objec-
tives may include specification of particular water
demand and operational modes. Based on these
2-6
-------
A Reference Guide for Utilities
specifications, a series of scenarios can be defined and
the model applied appropriately. Some software
products contain a scenario manager that helps the
user to define and manage a large number of specific
model runs. Additional scenarios can be developed in
order to test the sensitivity of the system to variations
in model parameters that are not known with certainty.
2.1.4.4 Analysis and Display of Results
Water distribution system models generate a large
amount of output. The amount of calculated informa-
tion increases with increasing model size and, for
EPS, the duration of the model run. Modern water
distribution system analysis software typically
provides a range of graphical and tabular displays
that help the user wade through the large amount of
output data so that it may be efficiently analyzed.
Figures 2-3, 2-4, and 2-5 contain examples of various
graphical and tabular outputs generated by the
EPANET software. These outputs represent a small
subset of types of graphics generated by most
modeling software. The output should be analyzed to
ensure that the model is operating properly and to
extract the information required in order to analyze
the specific problem being studied.
2.2 Modeling Water Quality In
Distribution System Networks
Water quality models use the output of hydraulic
models in conjunction with additional inputs
(described later in this section) to predict the temporal
and spatial distribution of a variety of constituents
within a distribution system. These constituents
include:
• The fraction of water originating from a
particular source.
• The age of water (e.g., duration since leaving
the source).
• The concentration of a non-reactive constituent
or tracer compound either added to or removed
from the system (e.g., chloride or fluoride).
• The concentration of a reactive compound
including the concentration of a secondary
disinfectant with additional input of its loss rate
(e.g., chlorine or chloramines) and the
concentration of disinfection by-products with
their growth rate (e.g., THMs).
The following subsection provides a brief history of
water quality modeling, an overview of theoretical
concepts related to water quality modeling, basic
model inputs, and model application.
2.2.1 History of Water Quality Modeling
The use of models to determine the spatial pattern of
Figure 2-3. EPANET Graphical Output Showing
Flow and Pressure.
Flow for Link 109
12
Time (hours)
Pressure for Node 111
Figure 2-4. Sample EPANET Time Series Plots of Flow,
Pressure, and Tank Water Level.
2-7
-------
A Reference Guide for Utilities
[&EPANET 2 - Net3.net - [Netwo... fT|[n"| X
H] File Edit View Project Report Window Help
n E; H « * x ** g
Node ID
JundO
JundB
Junc20
Junc35
Junc40
JuncSO
JuncGO
Junc601
Junc61
Junc101
Junc103
Junc105
Junc107
Junc109
Junc111
Elevati on
ft
147
32
129
12.5
131.9
^^^^^
0
0
0
?{j is • if |T j±i s *j> ®. e
Demand
GPM
0.00
0.00
0.00
1856.00
olo
^m
0.00
0.00
0.00
42 205.15
43 143.86
28.5
146.20
22 1 59.01
20.3 249.91
Head
ft
242.00
162.39
163.85
154.07
154.10
^^^^^
215.90
215.89
215.89
180.25
177.90
169.10
168.41
172.06
10 153.30 163.27
Pressure
psi
41.17
56.50
15.10
61.34
162
•n
93.55
93.55
93.55
59.90
58.45
60.92
63.41
65.76
66.41
A
V
Auto-Length Off GPM jf\ 100% XV: 1 1 .96, 30.91
[& EPANET2-Net3.net -[Netwo... _ \H\{
x
HH File Edit View Project Report Window Help - s x
Di*Bastex**0?{jgiiiff'* a ^ <±<
Link ID
Pipe 20
Pipe 40
Pipe 50
Pipe SO
Pipe 101
Pipe 103
Pipe 105
Pipe 107
Pipe 103
Pipe 111
Pipe 112
Pipe 113
Pipe 114
Pipe 115
Pipe 116
Auto-Length Off GF
Flow
GPM
13.66
-72.20
7731.31
3289.92
1739.64
1315.14
404.21
1595.79
1345.87
213.68
681.37
222.61
345.19
882.37
Velocity
fps
Unit
Heedless
Friction
Factor
0.03 0.00 0.125
0.00 0.00 0.000
0.00
5.48
4.15
2.78
3.82
1.15
2.55
0.00
3.33
4.35
1.74
4.39
0.17
1.18
0.000
0.014
0.024
0.019
0.019
0.023
0.020
3.82 4.39 0.019
0.61 0.15 0.025
1.93
1.25
0.021
1.42 1.13 0.024
2.20 2.55 0.023
2.50 2.01 0.021
A
V
^M jf\ 100% XV: 1 2.G7, 32.G1
Figure 2-5. EPANET Sample Tabular Outputs
(at time W.OOhrs).
water quality in a distribution system resulting from
sources of differing quality was suggested by Wood
(1980b) in a study of slurry flow in a pipe network.
The steady-state hydraulic model was extended by
solving a series of simultaneous equations at each
node. In a generalization of this formulation, Males
et al., (1985) used simultaneous equations to calcu-
late the spatial distribution of variables that could be
The ability to model the transport and fate of the water
constituents in a distribution system can help utility
managers perform a variety of water quality studies.
Examples include:
• Locating and sizing storage tanks and modifying
system operation to reduce water age.
• Modifying system design and operation to
provide a desired blend of waters from different
sources.
• Finding the best combination of: i) pipe
replacement, relining, and cleaning; ii) reduction
in storage holding time; iii) location and
injection rate of booster stations to maintain
desired disinfectant levels throughout the
system.
• Assessing and minimizing the risk of consumer
exposure to disinfectant by-products.
• Assessing system vulnerability to incidents of
external contamination.
• Designing a cost-efficient, routine monitoring
program to identify water quality variations and
potential problems.
associated with links and nodes such as concentra-
tion, travel times, costs, and others. This model,
called SOLVER, was a component of the Water
Supply Simulation Model (WSSM), an integrated data
base management, modeling, and display system that
was used to model water quality in networks (Clark
and Males, 1986). A more general "marching out"
solution was proposed by Males et al., (1988).
Although steady-state water quality models provided
some general understanding of water quality behavior
in distribution systems, the need for models that
would represent contaminant dynamics was recog-
nized. This resulted in the introduction of three such
dynamic models in the mid-1980s (Clark et al., 1986;
Liou and Kroon, 1986; and Hart et al., 1986).
The history and proliferation of water quality model-
ing in distribution systems can be traced back to two
expert workshops that were convened in 1991 and in
2003. The results of these workshops are presented in
AWWARF/USEPA (1991) and Powell et al., (2004).
Figure 2-6 illustrates the evolution of hydraulic and
water quality models since the 1930s.
2.2.2 Theoretical Concepts for Water Quality
Modeling
Various water quality processes are occurring in water
distribution systems that can lead to introduction of
contaminants and water quality transformations (see
Figure 1-2, presented earlier in Chapter 1) as water
moves through the distribution system. Cross
connections, failures at the treatment barrier, and
2-8
-------
A Reference Guide for Utilities
cuin
•="•53
Integration with GIS, SCADA
2ggg • • -optimization techniques
iggo H^ Powerful PC-based models
•7990 4Water quality models
7970 I
7960 ! ^Advent of computer models
7950 !
7940 I
<
M930 < Hardy Cross Method
Figure 2-6. Illustration of the Evolution of Hydraulic
and Water Quality Models.
transformations in the bulk phase can all degrade
water quality. Corrosion, leaching of pipe material,
biofilm formation, and scour can occur at the pipe
wall to degrade water quality. Bacteriological quality
changes may cause aesthetic problems involving taste
and odor development, discolored water, and other
adverse impacts.
In addition to the basic hydraulic modeling equations
presented earlier in this chapter, the water quality
models utilize various mathematical equations that
are based on conservation of constituent mass. These
models represent the following phenomena occurring
in a distribution system (Rossman et al., 2000):
• Advective transport of mass within pipes: A
dissolved substance will travel down the length
of a pipe with the same average velocity as the
carrier fluid while at the same time reacting
(either growing or decaying) at some given rate.
Longitudinal dispersion is not an important
transport mechanism in turbulent flow, which is
normal inside transmission mains under most
operating conditions. It may, however, be an
important factor in dead-end pipes or in low and
intermittent flow scenarios.
• Mixing of mass at pipe junctions: All water
quality models assume that, at junctions
receiving inflow from two or more pipes, the
mixing of fluid is complete and instantaneous.
Thus, the concentration of a substance in water
leaving the junction is simply the flow-
weighted sum of the concentrations in the
inflowing pipes.
• Mixing of mass within storage tanks: Most
water quality models assume that the contents
of storage tanks are completely mixed. See the
discussion in Section 2.4.1 for further details
and alternative representations.
• Reactions within pipes and storage tanks:
While a substance moves down a pipe or resides
in storage, it can undergo reaction. The rate of
reaction, measured in mass reacted per volume
of water per unit of time, will depend on the
type of water quality constituent being
modeled. Some constituents, such as fluoride,
do not react and are termed "conservative."
Other constituents, such as chlorine residual,
decay with time; while the generation of DBFs,
such as THMs, may increase over time. Some
constituents, such as chlorine, will react with
materials both in the bulk liquid phase and at
the liquid-pipe wall boundary.
Water quality models represent these phenomena
(transport within pipes, mixing at junctions and
storage tanks, and reaction kinetics in the bulk liquid
phase and at the liquid-pipe wall boundary) with a set
of mathematical equations. These equations are then
solved under an appropriate set of boundary and
initial conditions to predict the variation of water
quality throughout the distribution system.
Several solution methods are available for dynamic
water quality models (Rossman and Boulos, 1996).
All of these methods require that a hydraulic analysis
be run first to determine how flow quantities and
directions change from one time period to another
throughout the pipe network. The water quality
constituent is subsequently routed through each pipe
link and then mixed at downstream nodes with other
inflows into the node. For non-conservative sub-
stances, concentrations are continuously adjusted to
accommodate the decay or growth of the constituent
with time. This concentration is then released from
the node into its out-flowing pipes. This process
continues for all pipes and for the duration of the
simulation.
The methods described above are also applied when
modeling water age and source-tracing in water
quality models. Water age is equivalent to modeling
a reactive constituent that ages and combines linearly.
For example, for every hour that a "packet" of water
spends in a tank, its age will increase by one hour.
Additionally, combining a volume of water that is
four days old with a similar volume of water that is
eight days old will result in an average age of six
days. When modeling the fraction of water coming
from a designated source (source tracing), this
parameter is modeled as a conservative substance and
is linearly combined. For example, combining a
volume of water that is entirely from the designated
source with a similar volume of water from a different
2-9
-------
A Reference Guide for Utilities
Modeling the movement of a contaminant within the
distribution systems as it moves through the system
from various points of entry (e.g., wells or treatment
plants) to water users is based on three principles:
• Conservation of mass within differential lengths
of pipe.
• Complete and instantaneous mixing of the water
entering pipe junctions.
• Appropriate kinetic expressions for the growth or
decay of the substance as it flows through pipes
and storage facilities.
This change in concentration can be expressed by the
following differential equation:
(Equation 2-5)
dC SC
—! = -v..—i + k-C-
dt " <9x " "
where
Cjj = substance concentration (mg/L) at position x and
time t in the link between nodes i and j.
Vj = flow velocity in the link (equal to the link's flow rate
divided by its cross - sectional area) (m/sec)
kjj — rate at which the substance reacts within the link (sec"1)
According to Equation 2-5, the rate at which the mass of
material changes within a small section of pipe equals
the difference in mass flow into and out of the section
plus the rate of reaction within the section. It is as-
sumed that the velocities in the links are known
beforehand from the solution to a hydraulic model of
the network. In order to solve Equation 2-5, one needs
to know C.. at x=0 for all times (a boundary condition)
and a value for k...
ij
Equation 2-6 represents the concentration of material
leaving the junction and entering a pipe:
C,J@)i=o = " v n (Equation2-6)
where
C,.@x=0 = the concentration at the start of the link
connecting node i to node] in mg/L (i.e.,where x=0)
C tJ@x=L = the concentration at the end of a link, in mg/L
Q tj = flow from k to i
Equation 2-6 states that the concentration leaving a
junction equals the total mass of a substance flowing
into the junction divided by the total flow into the
junction.
source will provide a mixed volume calculated as 50
percent from the designated source.
2.2.3 Water Quality Model Inputs and
Application
In addition to the basic hydraulic model inputs
described in Section 2.1.3, the water quality models
require the following data elements to simulate the
behavior in a distribution system:
• Water Quality Boundary Conditions - A water
quality model requires the quality of all
external inflows to the network and the water
quality throughout the network be specified at
the start of the simulation. Data on external
inflows can be obtained from existing source
monitoring records when simulating existing
operations or could be set to desired values to
investigate operational changes. Initial water
quality values can be estimated based on field
data. Alternatively, best estimates can be made
for initial conditions. Then the model is run for
a sufficiently long period of time under a
repeating pattern of source and demand inputs
so that the initial conditions, especially in
storage tanks, do not influence the water quality
predictions in the distribution system. The
water age and source tracing options only
require input from the hydraulic model.
• Reaction Rate Data - For non-conservative
substances, information is needed on how the
constituents decay or grow over time.
Modeling the fate of a residual disinfectant is
one of the most common applications of
network water quality models. The two most
frequently used disinfectants in distribution
systems are chlorine and chloramines (a reactant
of chlorine and ammonia). Free chlorine is
more reactive than chloramine and its reaction
kinetics have been studied more extensively.
Studies have shown that there are two separate
reaction mechanisms for chlorine decay, one
involving reactions within the bulk fluid and
another involving reactions with material on or
released from the pipe wall (Vasconcelos et al.,
1997). Bulk decay is typically represented as a
first order exponential decay function with a
single decay coefficient specified to represent
the decay over time. In some circumstances, this
function does not adequately represent the
observed decay characteristics, and more
complex formulations may be used to describe
the decay. Wall reaction represents the
disinfectant decay due to contact with
oxidizeable substances at the pipe wall, such as
corrosion products or biofilm. The most widely
used approach for representing wall demand
considers two interacting processes - transport
2-10
-------
A Reference Guide for Utilities
Storage tanks are usually modeled as completely
mixed, variable volume reactors in which the changes
in volume and concentration over time are as follows:
dVs
~dT
(Equation 2-7)
at
(Equation 2-8)
where
Cs = the concentration for tanks, in mg/L
dt = change in time, in seconds
Qts = flow from node k to s, in ftVsec (mVsec)
Qsj - flow from node s to j, in ftVsec (mVsec)
dVs - change in volume of tank at nodes, in ft3(m3)
V = volume of tank at nodes, in ft3(m3)
Cts - concentration of contaminant in link ks, in mg/ft3 (mg/m3)
ka = decay coeficient between nodes i and j, in sec"1
Many algorithms and methods exist for the numerical
solution of fluid flows described by the Navier-Stokes
equations. These algorithms can be classified as
Eulerian or Lagrangian and as either time-driven or
event-driven. In a Eulerian method, the movement of
the fluid is viewed from a stationary grid as the water
moves through the system. On the contrary, in a
Lagrangian method, the analysis is viewed from a
framework that is moving with the flow. Time-driven
methods assess the system at fixed time steps. Event-
driven methods evaluate the system only when there is
a discrete change in water quality such as a pulse of
water with different concentrations entering or leaving
a pipe. Various methodologies combine either Eulerian
or Lagrangian solutions (or hybrid combinations of
these two cases) with either time-driven or event-driven
procedures.
of the disinfectant from the bulk flow to the
wall and interaction with the wall (Rossman et
al., 1994). Recent studies have suggested that
this formulation may not adequately represent
the actual wall demand processes and that
further research is needed (Clark et al., 2005;
Grayman et al., 2002; DiGiano and Zhang,
2004). There has been little study on the nature
of the wall reaction in chloraminated systems.
A limited amount of modeling of the growth of
DBFs (most notably THMs) has been performed
assuming an exponential growth approaching a
maximum value corresponding to the THM
formation potential. Both the formation
potential and the growth rate constant must be
specified in this type of model (Clark et al.,
1996). There has been extensive research on
biofilm formation in distribution systems and
this has led to the development of several
theoretical models of this phenomenon (Powell
et al., 2004). However, these models are
generally quite complex involving many
parameters that are difficult to determine, and
thus are not ready for inclusion in a general
water distribution system model.
The following section provides an overview of
available software for hydraulic and water quality
modeling.
Distribution system water quality models are generally
limited to tracking the dynamics of a single component
(e.g., chlorine, water age) at a time when the selected
component is transported throughout the network of
pipes and storage tanks. Such models do not consider
interactions between different components in the
flowing water or complex reactions between compo-
nents that are transported with the water and surface
components that are fixed to the pipe wall. This can be a
significant limitation when modeling reactive compo-
nents, for example when chlorine residual is modeled for
a case where there are multiple sources with significant
differences in water quality characteristics. Another
more complex example that is not adequately repre-
sented by the single-species model is modeling of DBF
formation. A solution to this deficiency is a general-
purpose, multi-species capability that is being added to
EPANET (Uber et al., 2004). This addition will allow
users to program their own chemical/physical/biological
reactions in EPANET with almost unlimited interaction
capability between multiple species.
2.3 Hydraulic and Water Quality
Modeling Software
A variety of software packages are available to
perform hydraulic and water quality modeling. A
majority of these packages utilize the EPANET
formulation as the basic computation engine. A full
discussion of individual software packages is beyond
the scope of this document. The following subsec-
tions briefly describe the EPANET model and
summarize the features of other available software.
2.3.1 EPANET Software
EPANET was initially developed in 1993 as a
distribution system hydraulic-water quality model to
support research efforts at EPA (Rossman et al., 1994).
The development of the EPANET software has also
satisfied the need for a comprehensive public-sector
model and has served as the hydraulic and water
quality "engine" for many commercial models.
2-11
-------
A Reference Guide for Utilities
EPANET can be used for both steady-state and EPS
hydraulic simulations. In addition, it is designed to
be a research tool for modeling the movement and fate
of drinking water constituents within distribution
systems. EPANET can be operated in the SI (metric)
or British systems of measurement.
The water quality routines in EPANET can be used to
model concentrations of reactive and conservative
substances, changes in age of water and travel time to
a node, and the percentage of water reaching any node
from any other node. Outputs from EPANET include:
• color-coded network maps,
• time series plots, and
• tabular reports.
Example outputs from EPANET were previously
presented in Figures 2-3, 2-4, and 2-5.
2.3.2 Commercial Hydraulic-Water Quality
Modeling Software
In addition to EPANET, there are several commercial
software packages that are widely used in the U.S. and
internationally. Most of these packages are based on
the EPANET formulation and include value-added
components as parts of GUI that increase the capabil-
ity of the software. Examples of such value-added
components that are part of one or more of the
commercially available software packages include:
• Scenario manager: Manage inputs and outputs
of a group of model runs.
• Calibration optimization: Utilize genetic
algorithm optimization technique to determine
model parameters that best fit a set of field data.
• Design optimization: Utilize genetic algorithm
optimization techniques to select pipe sizes that
minimize costs or other selected objectives.
• Integration with GIS or CAD: Water distribution
model directly integrates with GIS or CAD to
assist in constructing or updating model and
In addition to the standard use of EPANET in a
Windows environment using the graphical user
interface (GUI), the functionality of EPANET can be
accessed through the EPANET toolkit. The toolkit is a
series of open source routines available in both Visual
Basic and C (programming language) that can be used
as is or modified and accessed from a user's own
computer program. This powerful capability has been
widely used throughout the world to support both
research and specific applications in the field of water
distribution system analysis.
display results.
• Flexible output graphics: Provides convenient
ways to modify parameters for graphical
displays of output data.
• Energy management: Calculates energy use for
a selected alternative.
• Automated fire-flow analysis: Assesses the
availability of fire flow at a range of nodes and
determines whether a system meets fire-flow
requirements.
• Water security and vulnerability assessment
methods, skeletonization, and demand
allocation tools.
Table 2-1 provides a summary listing of major
commercial software and a Web link where additional
details may be obtained on specific features and
current version availability/pricing.
2.4 Additional Modeling Tools
In addition to standard hydraulic and water quality
modeling of distribution systems, there are several
other related types of models that can be used to
assess hydraulic and water quality behavior in
distribution systems. These include: storage modeling
tools, transient (water hammer) modeling tools,
optimization tools, and probabilistic models. Each of
these types of models are briefly described and
demonstrated in the following sections.
2.4.1 Storage Modeling Tools
An important aspect of water quality and contaminant
propagation in drinking water distribution systems is
the effect of system storage. Most utilities use some
type of ground or elevated storage system to process
water during time periods when treatment facilities
would otherwise be idle. It is then possible to
distribute and store water at one or more locations in
the service area closest to the user.
The principal advantage of distribution storage is that
it equalizes demands on supply sources, production
works, and transmission and distribution mains. As a
result, the sizes or capacities of these elements may be
minimized and peak power tariff periods for pumping
can often be avoided. Additionally, system flows and
pressures are improved and stabilized to better serve
the customers throughout the service area. Finally,
reserve supplies are provided in the distribution
system for emergencies, such as fire fighting and
power outages.
In most municipal water systems, less than 25 percent
of the volume of the storage in tanks is actively used
(on a daily basis) under routine conditions. As the
2-12
-------
A Reference Guide for Utilities
Table 2-1. Available Hydraulic and Water Quality Network Modeling Software Packages
Network Modeling Software
AQUIS
EPANET
InfoWater H2ONET/H2OMAP
Info Works WS
MikeNet
Pipe2000
PipelineNet
SynerGEE Water
WaterCAD/WaterGEMS
STANET
Wadiso
Company
Seven Technologies
EPA
MWHSoft
Wallingford Software
DHI, Boss International
University of Kentucky
SAIC, TSWG
Advantica
Haestad Methods
Fisher-Uhrig Engineering
GLS Eng. Software
EPANET
Based
X
X
X
X
X
X
Website
www.7t.dk/aquis
www.epa.gov/ord/nrmrl/wswrd/epanet.html
www.mwhsoft.com
www.wallingfordsoftware.com
www.dhisoftware.com/mikenet
www.kypipe.com
www.tswg.gov/tswg/ip/pipelinenettb.htm
www.advantica.biz
www.haestad.com
www.stanet.net
www.wadiso.com
water level drops, tank controls require high-service
pumps to start in order to satisfy demand and refilling
of the tanks. The remaining water in the tanks (70 to
75 percent) is normally held in reserve as dedicated
fire or emergency storage. This water tends to be
stagnant and may cause water quality problems.
Storage tanks and reservoirs are the most visible
components of a water distribution system, but are
often the least understood in terms of their effect 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 unwanted complex
chemical and biological changes that may result in
the deterioration of water quality. These storage tanks
and reservoirs also contribute to increased residence
time in drinking water systems. This increased
residence time can contribute to the loss of disinfec-
tant residuals and cause subsequent growth of
microorganisms. Modeling can provide information
on what will happen in existing, modified or proposed
distribution system tanks and reservoirs under a range
of operating situations (Grayman et al., 2004a).
Three primary types of models are used for represent-
ing storage tanks and reservoirs: computational fluid
dynamics (CFD) models, compartment models, and
physical scale models. In mathematical models,
equations are written to simulate the behavior of
water in a tank or reservoir. These models range from
detailed representations of the hydraulic mixing
phenomena in the facility called CFD models to
simplified conceptual representations of the mixing
behavior called compartment or systems models.
Physical scale models are constructed from materials
such as wood or plastic. Dyes or chemicals are used
to trace the movement of water through the model.
2.4.1.1 CFD Models
CFD models use mathematical equations to simulate
flow patterns, heat transfer, and chemical reactions.
Partial differential equations representing conserva-
tion of mass, momentum, and energy are solved
numerically for a two- or three-dimensional grid that
approximates the geometry of the tank. CFD model-
ing has been used widely in the chemical, nuclear, and
mechanical engineering fields, and in recent years has
emerged as a modeling tool in the drinking water
industry (Grayman and Arnold, 2003). CFD models
can be used to simulate temperature variations,
unsteady hydraulic and water quality conditions, and
decay of constituents in storage facilities. Signifi-
cant experience is required to apply CFD models, and
model run times of many hours, days, or even weeks
are required for complex situations. Figure 2-7 depicts
a graphical output from a CFD model showing the
concentration throughout a tank at a snapshot in time
resulting from a tracer that has been injected into the
inflow.
Many generalized CFD software packages are
available that can be used to construct CFD models of
tanks. Examples of such packages are listed in Table
2-2. These packages vary in terms of capabilities,
solution methods, ease of use, and support. Prior to
selection of a package, the specific needs and
capabilities of the user should be carefully evaluated.
2-13
-------
A Reference Guide for Utilities
Negative Buoyancy
Neutral Buoyancy
Rostov* Buoyancy
Figure 2-7. Graphical Output from a CFD Model
Showing Tracer Concentration in a Tank.
Generally the purchase or lease of these packages is
significant (typically on the order of $25,000 per
year) and significant training/expertise is required to
effectively apply them.
2.4.1.2 Compartment Models
Compartment models are a class of models in which
physical processes (i.e., the mixing phenomena in the
tank or reservoir) are represented by highly concep-
tual, empirical relationships. This type of model is
also referred to as a black box model, or input-output
model. Since such models do not use detailed
mathematical equations to describe the movement of
water within the tank, they rely on engineering
judgment or upon field data and past experience to
define the parameters that control the behavior of the
model. Compartment models are used in water
distribution network models to represent mixing in
tanks and reservoirs. Various assumptions can be
made in these models about the mixing behavior in
tanks including complete and instantaneous mixing,
plug flow, last-in/first-out (LIFO) behavior, and multi-
Table 2-2. Example CFD Modeling Software Packages
CFD Package
CFD-ACE
Cfdesign
CFX
FLOW-3D
Fluent
Phoenics
SWIFT
Sinda/Fluint
PAB3D
Company
CFD Research Corp.
Blue Ridge Numerics
Ansys, Inc.
Flow Science, Inc.
Fluent, Inc.
CHAM
AVL
C&R Technologies
Analytical Services &
Materials
Website
www.cfdrc.com
www.brni.com
www.software.aeat.com/cfx
www.flow3d.com
www.fluent.com
www.cham.co.uk
www.avl.com
www.crtech.com
www.asm-usa.com
compartment models. Both conservative substances
and substances that decay according to a first-order
decay function may be simulated in addition to
simulation of water age. Compartment models are
relatively easy to use and run in seconds as opposed
to the long run times of CFD models.
Compartment models of tanks are available as part of
most water distribution system models. EPANET and
several of its derivative commercial models allow the
user to select from four options - a complete mix
model, a plug flow first-in/first-out (FIFO) model, a
LIFO (short circuiting) model and a two-compartment
model. A stand-alone model called CompTank
provides a wide range of alternatives and allows the
user to model water age and reactive or conservative
substances over a long period of time (Grayman et al.,
2000). This model uses tank inflow and outflow
information that is generally available from SCADA
records as its primary input.
2.4.1.3 Physical Scale Models
Physical scale models provide a relatively inexpen-
sive mechanism for studying the mixing characteris-
tics of tanks. In a physical scale model, a tracer
chemical is added to the inflow (or internally within
the model) and the movement of the tracer is moni-
tored during the experiment (Grayman et al., 2000).
Tracer substances include visible dyes, which are
appropriate for developing a qualitative understand-
ing of mixing behavior, and chemicals (e.g., calcium
chloride) that can be measured by sensors in the tanks
and used for quantitative assessments. Use of tracers
of different density or careful control of temperature
of the tracer can be used to study the impacts of
thermal variations on mixing. Laws of similitude in
hydraulics must be followed in order to account for
the scaling effects. Scale models can vary in size and
complexity from small tabletop models to large-scale
models built in hydraulics laboratories. Figure 2-8
depicts such a large-scale model.
Figure 2-8. A Large Physical Model of a Tank (Source:
Bureau of Reclamation Laboratory).
2-14
-------
A Reference Guide for Utilities
In an advanced technology form of physical scale
modeling, three-dimensional laser induced fluores-
cence is being used to provide detailed measurements
of mixing in tanks (Roberts and Tian, 2002). Figure 2-
9 shows an illustration of output from this technology.
2.4.2 Transient Analysis Software
A hydraulic transient is a rapid change in pressure
associated with a pressure wave that moves rapidly
through a piping system. A transient can be caused
by a variety of events, such as rapid operation of a
valve (including fire hydrants) or rapid pump starts
and stops. If the magnitude of the resulting
pressure wave is large enough and adequate
transient control measures are not in place, a
transient can cause a water hammer leading to
failure of hydraulic components. It can also lead to
instantaneous low or negative pressures that can
result in intrusion of untreated water into the pipe,
potentially resulting in contamination. Transient
events are highly dynamic and sophisticated.
Mathematical models are required to analyze their
movement in a distribution system.
Several commercial software packages for performing
transient analysis in water distribution systems are
available. Examples of such software are listed in
Table 2-3. The technical capabilities, user interface,
solution methods, graphical display, and technical
support and training vary considerably among the
packages.
2.4.3 Optimization Tools
Optimization tools allow the user to evaluate a large
number of options and to select the specific alterna-
tive that gives the best results in terms of predefined
objective functions. In the area of water distribution
system analysis, optimization models are used for
calibration, design, and operational purposes. These
applications are briefly described in the following
subsections.
2.4.3.1 Optimizing Calibration
Calibration of a water distribution system model
involves adjustments in various model parameters so
that the model agrees with field measurements of flow
and pressure. Such a tool is used most frequently with
flow and pressure measurements taken during flow
(hydrant) tests to stress the system. Parameters that
are typically adjusted include roughness factors,
demands, and status of isolation valves.
B 13 16 IS 23 Z7 3D
ft
.20 -15 .10 .5 0 5 to 15 20 25
x(cia)
Figure 2-9. Graphical Output Based on 3-D Laser Induced Fluorescence with a Physical Scale Model Showing
Mixing in Tank (Source: Georgia Tech).
Table 2-3. Example Transient Modeling Software Packages
Transient Modeling Software
AQUIS Surge
HAMMER
Hytran v3.0
Impulse
InfoSurge, H20Surge
Company
Seven Technologies
Haestad Methods
Hytran Solutions
Applied Flow Technology
MWHSoft
Website
www.7t.dk/aquis
www.haestad.com
www.hytran.net
www.aft.com/products/impulse
www.mwhsoft.com
2-15
-------
A Reference Guide for Utilities
The production of transient low-and negative-
pressures in otherwise pressurized drinking water
supply distribution systems creates the opportunity
for contaminated water to enter the pipe from
outside. Such events may be caused by the sudden
shutdown of pumps or by other operational events
such as flushing, hydrant use, and main breaks.
Figure 2-10 illustrates an event that results in a
negative pressure transient for 22 seconds caused
by a power outage associated with a lightning
strike.
In a series of research projects (LeChevallier et al.,
2003; Gullick et al., 2004), the frequency and
location of low-and negative-pressures in represen-
tative distribution systems were measured under
normal operating conditions and during specific
operational events. These investigators also
confirmed that fecal indicators and culturable
human viruses were present in the soil and water
exterior to the distribution system pipes. Their
research shows that a well-calibrated hydraulic
surge model can be used to simulate the occurrence
of pressure transients under a variety of operational
scenarios, and a model can also be used to deter-
mine optimal mitigation measures.
Although there are insufficient data to indicate
whether pressure transients pose a substantial risk
to water quality in the distribution system, mitiga-
tion techniques can be implemented. These
techniques include the maintenance of an effective
disinfectant residual throughout the distribution
system, leak control, redesign of air relief venting,
installation of hydro-pneumatic tanks, and more
rigorous application of existing engineering
standards.
negative 4.4 psi
± 2.0 psi)
for 20 seconds
Figure 2-10. Negative Pressure Transient Associated
with a Power Outage.
Use of manual adjustment techniques may involve
many tedious runs of a distribution system model
until the resulting predicted flows and pressures
approximate the values observed in the field. When
an optimization model is applied, the user defines an
objective function, such as minimizing the square of
the difference between observed and predicted values
(for pressure and flow). The optimization algorithm
then uses some type of controlled search method to
identify the set of model parameters that will result in
the best results (i.e., minimize the error). The user will
generally set constraints on parameters so that the
resulting values are reasonable. For example, the user
may specify that the allowable range for the rough-
ness factor for a certain set of ductile iron pipes range
between 90 and 120.
Over the past 40 years, various techniques have been
applied as part of automated calibration methods
(Rahal et al., 1980; Walski et al., 2003). The most
common optimization technique in use today couples
a hydraulic model with an optimization routine using
genetic algorithms. Genetic algorithms are based on
the theory of genetics in which successive population
trials are generated with the "fittest" ones surviving
to breed and evolve into increasingly desirable
offspring solutions. The fitness of a solution is based
on the objective functions that were previously
described. Genetic algorithm-based calibration tools
are available as optional components of several water
distribution system analysis software packages.
2.4.3.2 Design Optimization
In a manner analogous to the calibration optimization
technique described above, design optimization
techniques evaluate a large number of distribution
system design options and select the one that
provides the best solution (Lansey, 2000). Schaake
and Lai (1969) first proposed such an approach and
applied it to the design of major transmission lines
providing water to New York City. Since that time,
numerous papers have been written on the subject
(Walski et al., 2003) and have included a variety of
techniques such as linear programming, dynamic
programming, mixed integer programming, heuristic
algorithms, gradient search methods, enumeration
methods, genetic algorithms, and simulated anneal-
ing. In recent years, genetic algorithm methods have
been favored for this problem and have been widely
used in a variety of situations and are included in
several commercial software packages. The user
should, however, be aware that genetic algorithms do
not guarantee optimality. These algorithms must be
run several times to ensure near optimal solutions.
Typically, design optimization tools limit a user to
choose from designated piping options and to size
the pipes to meet present and future demands. Cost
minimization is the most common objective function.
2-16
-------
Additionally, some researchers have incorporated
reliability and capacity considerations (Mays, 1989).
2.4.3.3 Optimization of Operation
Models can also be used to optimize operations of a
distribution system (Goldman et al., 2000). The most
common areas of operation where such models have
been applied are in energy management and water
quality. Chase et al. (1994) describe a computer
program to control energy costs that incorporates a
hydraulic model, a pump optimization program, and
an interface. In the water quality area, Uber et al.
(2003) used optimization techniques to determine
optimal location and operation of chlorine booster
stations. Jentgen et al. (2003) implemented a proto-
type energy and water quality management system at
Colorado Springs Utilities. This system combines a
simplified distribution system model and an optimi-
zation routine to adjust operation of the water system
and power generation system in near real-time.
2.4.4 Probabilistic Models
Hydraulic and water quality models of distribution
systems are deterministic models. For a set of network
parameters and specific operations and demands, the
model produces a single set of resulting flows and
pressures. However, there is uncertainty in many of
the aspects of these models including parameters such
as roughness, demands, actual inside diameter of
pipes, valve settings, and system controls. This
uncertainty is generally due to both imperfect
knowledge and natural variability. An emerging
procedure is to embed a deterministic network model
within a probabilistic framework and to examine the
effect of uncertainty on the results.
The most common approach to incorporating uncer-
tainty in models is the use of a Monte Carlo simula-
tion (Vose, 2000). In this method, probability
distributions are assigned to model parameters to
represent the uncertainty associated with each
parameter. The distribution system model is then run
many times with parameter values being randomly
drawn from the probability distributions. The results
of many iterations are combined to determine the
most likely result and a distribution of results. This
approach has been used in legal cases where historical
contamination events have been reconstructed
(Grayman et al., 2004b), in evaluation of the impacts
of purposeful contamination (Murray et al., 2004) and
modeling bacterial regrowth in distribution systems
(DiGiano and Zhang, 2004).
2.5 Summary and Conclusions
Acquiring and utilizing the proper data is very
important for implementing water distribution system
models. The key inputs include the characterization
A Reference Guide for Utilities
of the pipe network (e.g., pipes, pumps, tanks, and
valves), water-demand information (temporal varia-
tions are required in EPS), topographic information
(elevations assigned to nodes), control information
that describes how the system is operated, and EPS
solution parameters (e.g., time steps, tolerances as
required by the solution techniques). Periodic
calibration and validation of a model is important to
achieve optimum results.
Models have become widely accepted within the
water utility industry as a mechanism to simulate the
hydraulic and water quality behavior of a real or
proposed distribution system. They are routinely
used for a number of tasks including capital invest-
ment decisions, master plan development, and fire
protection capacity design. Furthermore, these
models have become very sophisticated and typically
simulate both hydraulic and water quality behavior.
Many modeling packages are integrated with GIS or
CAD. Some software packages incorporate water
hammers and tank mixing. EPANET is a public sector
hydraulic/water quality model developed by EPA.
EPANET also serves as the computation engine for
many of the commercial models used by water
utilities throughout the country. In addition to
EPANET and EPANET-based water distribution
system models, there are several other tools available
to users for studying specific needs, such as transient
analysis and optimization analysis.
To successfully apply a model to study a problem, one
should clearly define the objectives and select an
appropriate tool. Thereafter, understanding the
accuracy of the input data and limitations of the
model will enable the user to better interpret the
results of the analysis and develop appropriate
solutions.
2-17
-------
A Reference Guide for Utilities
Many of the assumptions and methodologies in use
today in water distribution system modeling date
back to the early work of Hardy Cross (1936). With
the monumental increase in computational power
and improvements in the ability to measure flow in
experimental distribution systems, it is natural that
some of the basic assumptions are being examined
and challenged. Three notable examples of active
research areas include the following:
• Distribution system water quality models
currently assume advective flow that results
in water quality pulses moving through a
pipe without spreading out longitudinally.
Lee and Buchberger (2001) have studied
pipe flow and found that dispersion has a
significant effect on concentration profiles,
especially in cases of intermittent laminar
flow. Lee (2004) developed an analytical
equation which describes the unsteady
dispersion of changing flow velocity in
pipes based on the classic one-dimensional
advection-dispersion equation by Taylor
(1953). Tzatchkov et al., (2002) have
developed an extension to the standard
EPANET model that includes dispersion.
• In distribution system models, deterministic
demands are assigned to nodes. Buchberger
et al., (2003) monitored water use at the
individual home and neighborhood level
and found that there are significant short-
term variations in water use. They have
developed a model that represents water
use as a series of pulses which can be
simulated using a Poisson Rectangular
Pulse model to capture the natural
variability associated with water use.
Distribution system models currently
assume complete mixing at a junction. As
a result, if there are two pipes with flow
entering the junction and two pipes
through which the flow exits, the chemical
content of the water in the two exiting
pipes will be identical and represent an
average of the characteristics of the two
entering pipes. Van Bloemen Waanders et
al., (2005) have tested this assumption
using both laboratory analysis and CFD
modeling. Figure 2-1 la depicts the
velocity field at a junction. Figure 2-lib
presents the corresponding tracer
concentrations at that junction. The figures
indicate that the complete mix assumption
would lead to some inaccuracy in
computing chemical transport in a
distribution system.
Figure 2-11 a. Velocity Field at a Junction.
Figure 2-1 lb. Tracer Concentration at a Junction.
2-18
-------
A Reference Guide for Utilities
References
American Water Works Association (AWWA). Com-
puter Modeling of Water Distribution Systems, AWWA
ManualM-32, Denver, CO. 2004.
AwwaRF/U.S. EPA. Water Quality Modeling in
Distribution Systems. Edited by W.M. Grayman, R.M.
Clark, and R.M. Males. AwwaRF-AWWA. Denver.
CO. 1991.
Buchberger, S.G., J.T. Carter, Y.H. Lee, and T.G.
Schade. "Random Demands, Travel Times, and Water
Quality in Deadends." AwwaRF/AWWA, Denver, CO.
2003.
Bureau of Reclamation Laboratory. Photo provided
by Bates Engineering, Inc., 2004.
Chase, D.V., D. Guastella, K.LLai, and L.E. Ormsbee.
"Energy Management for Large Scale Water Supply
Systems." AWWA Computer Conference, Los
Angeles, CA. 1994.
Clark, R.M., W.M. Grayman, R.M. Males, and J.A.
Coyle. "Predicting Water Quality in Distribution
Systems." Proceedings, AWWA Distribution System
Symposium, Minneapolis, MN. 1986.
Clark, R.M., and R.M. Males. "Developing and
Applying the Water Supply Simulation Model,"
Journal of AWWA, 78(8). 1986.
Clark, R.M., and W.M. Grayman. Modeling Water
Quality in Drinking Water Systems. AWWA. 1998.
Clark R.M., R.C. Haught, and W.M. Grayman.
"Predicting Chlorine Residuals in Unlined Metallic
Pipes." Proceedings, ASCE-EWRI World Water &
Environmental Resources Congress, Anchorage, AK.
2005.
Clark, R.M., H. Pourmaghaddas, L.G. Wymer, and R.C.
Dressman. "Modeling the Kinetics of Chlorination
By-Product Formation: The Effects of Bromide,"
Journal Water Supply Research and Technology -
Aqua, 45(1): 1-8. 1996.
Cross, H. "Analysis of Flow in Networks of Conduits
or Conductors," University of Illinois Engineering
Experiment Station Bulletin 286, Urbana, IL. 1936.
DiGiano, F. A., and W. Zhang. "Uncertainty Analysis
in a Mechanistic Model of Bacterial Regrowth in
Distribution System." Environmental Science and
Technology 38:22:5925-5931. 2004.
EPA. Stage 2 Disinfectants and Disinfection
Byproducts Initial Distribution System Evaluation
Guidance Manual, Draft. 2003.
Georgia Tech. Photo provided by Dr. Philip J. W.
Roberts, School of Civil Engineering, Georgia
Institute of Technology, 2004.
Goldman, F.E., A.B.A. Sakarya, L.E. Ormsbee, J.G.
Uber, and L.W Mays. "Optimization Models for
Operation." Chapter in Water Distribution System
Handbook. Edited by Larry Mays for McGraw-Hill,
NY. 2000.
Grayman, W.M., R.M. Clark, and R.M. Males.
"Modeling Distribution System Water Quality
Dynamic Approach," Journal of Water Resources
Planning and Management, ASCE, 114(3). 1988.
Grayman, W.M. "Put Your Hydraulic Model to Work
- How That Costly Model Can Earn its Keep by
Making Operations More Effective." Proceedings,
AWWA Annual Conference, Denver, CO. 2000.
Grayman, W.M.,L.A. Rossman, M.A. Gill, Y Li, and
D.E. Guastella. "Measuring and Modeling Disinfec-
tant Wall Demand in Metallic Pipes." Proceedings,
EWRI Conference on Water Resources Planning and
Management, Reston VA. ASCE. 2002.
Grayman, W.M., L.A. Rossman, R.A. Deininger, C.D.
Smith., C.N. Arnold., and J.F. Smith. "Mixing and
Aging of Water in Distribution System Storage
Facilities." Journal of AWWA, 96(9):70-80. 2004a.
Grayman W.M., R.M. Clark, B.L. Harding, M. Maslia,
and J. Aramini. "Reconstructing Historical Contami-
nation Events," Chapter in Water Supply Systems
Security. Edited by Larry Mays, McGraw-Hill, NY.
2004b.
Grayman, W.M., and C. N. Arnold. "Overview of CFD
Methods in Analysis of Distribution System Tanks
and Reservoirs." Proceedings, AWWA Annual
Conference, AWWA, Denver, CO. 2003.
Grayman, W.M., and L.A. Rossman. "Integrating a
Network Model with an Expert System." Proceedings,
AWWA Computer Conference, AWWA, Denver, CO.
1994.
Grayman, W.M., L.A. Rossman, C.N. Arnold, R.A.
Deininger, C.D. Smith, J.F. Smith, and R. Schnipke.
Water Quality Modeling of Distribution System
Storage Facilities. AwwaRF and AWWA. 2000.
2-19
-------
A Reference Guide for Utilities
Gullick, R.W., M.W. LeChevallier, R.C. Svindland,
and M. Friedman. "Occurrence of Transient Low and
Negative Pressures in Distribution Systems." Journal
ofAWWA. 96(11): 52-66. November 2004.
Hart, F.L., J.L. Meader, and S.N. Chiang. "CLNET A
Simulation Model for Tracing Chlorine Residuals in a
Potable Water Distribution Network." Proceedings,
AWWA Distribution System Symposium, Minneapo-
lis, MN. 1986.
Jentgen, L., S. Conrad, R. Riddle, E. Von Sacken, K.
Stone, W.M. Grayman, and S. Ranade. Implementing
a Prototype Energy and Water Quality Management
System. AwwaRF, Denver, CO. 2003.
Jeppson, R.W. Analysis of Flow in Pipe Networks.
Ann Arbor Science, Ann Arbor, MI. 1976.
Johnson K.E., and J. Loux. Water and Land Use:
Planning Wisely For California's Future. Solano
Press, Point Arena, CA. 2004.
Lansey, K.E. "Optimal Design of Water Distribution
Systems." Chapter in Water Distribution System
Handbook. Edited by Larry Mays for McGraw-Hill,
NY. 2000.
Larock, B.E., R.W. Jeppson, and G.Z. Watters. Hy-
draulics of Pipeline Systems. CRC Press, LLC, Boca
Raton, FL. pp 12-26. 2000.
LeChevallier, M.W., R.W. Gullick, M.R. Karim, M.
Friedman and J.E. Funk. "The Potential for Health
Risks from Intrusion of Contaminants into the
Distribution System from Pressure Transients,"
Journal of Water and Health, IWA Publishing, pp. 3-
14. January 2003.
Lee, Y. and S.G. Buchberger. "Estimation of disper-
sion in unsteady random flow conditions in dead-end
pipes of water distribution systems." Proceedings,
ASCE-EWRI World Water & Environmental Re-
sources Congress, Orlando, FL. 2001.
Lee, Y. Mass Dispersion in Intermittent Laminar Flow.
Ph.D. Dissertation in Civil and Environmental
Engineering, University of Cincinnati, Cincinnati,
Ohio. 2004.
Liou, C.P, and J.R. Kroon. "Propagation and Distri-
bution of Waterborne Substances in Networks,"
Proceedings, AWWA Distribution System Symposium,
Minneapolis, MN. 1986.
Males, R.M., R.M. Clark, P.J. Wehrman, and WE.
Gates. "Algorithm for Mixing Problems in Water
Systems," Journal Hydraulic Engineering, ASCE,
111(2). 1985.
Males, R.M., W.M. Grayman, R.M. Clark. "Modeling
Water Quality in Distribution Systems," Journal
Water Resources Planning and Management, ASCE,
114(2). 1988.
Mays, L.W., Editor. Reliability Analysis of Water
Distribution Systems. ASCE. NY. 1989.
Murray, R., R. Janke, and J. Uber. "The Threat
Ensemble Vulnerability Assessment (TEVA) Program
for Drinking Water Distribution System Security."
Proceedings, ASCE-EWRI World Water & Environ-
mental Resources Congress, Salt Lake City, UT. 2004.
Powell, J., J. Clement, M. Brandt, R. Casey, D. Holt,
W. Grayman, and M. LeChevallier. "Predictive
Models for Water Quality in Distribution Systems."
AwwaRF, Denver, CO. 2004.
Rahal, C.M., M.J.H. Sterling, and B. Coulbeck.
"Parameter Tuning for Simulation Methods of Water
Distribution Networks." Proceedings, Institute of
Civil Engineers, Part 269, 751. 1980.
Roberts, P.J.W, and X. Tian. "Application of Three-
Dimensional Laser-Induced Fluorescence to Stratified
Turbulent Mixing Processes." Proceedings, ASCE
Hydraulic Measurements and Experiments Confer-
ence. 2002.
Rossman, L.A., and PF. Boulos. "Numerical Methods
for Modeling Water Quality in Distribution Systems:
A Comparison." Journal Water Resources Planning
and Management, ASCE, 122(2):137-146. 1996.
Rossman, L.A. EPANET Version 2 Users Manual,
Drinking Water Research Division, EPA, Cincinnati,
OH. 2000.
Rossman, L.A., R.M. Clark, and W.M. Grayman.
"Modeling Chlorine Residuals in Drinking Water
Distribution Systems." Journal Environmental
Engineering, 120(4) :803-820. 1994.
Schaake, J., and D. Lai. Linear Programming and
Dynamic Programming Applications to Water
Distribution Network Design - Report 116. Depart-
ment of Civil Engineering, MIT, Cambridge, MA.
1969.
2-20
-------
A Reference Guide for Utilities
Taylor, G.I. "Dispersion of soluble matter in solvent
flowing slowly through a tube," Journal of the Royal
Society of London Series A, The Royal Society,
London, UK. 219:186-203. 1953.
Tzatchkov, V.G., A.A. Aldama, and F.I. Arreguin.
"Advection-dispersion Reaction Modeling in Water
Distribution Networks." Journal Water Resources
Planning & Management, 128(5):334-342. 2002.
Uber, J., F. Shang, M. Polycarpou, and Z.Wang.
"Feedback Control of Booster Chlorination Systems."
AwwaRF, Denver, CO. 2003.
Uber, J., F. Shang, and L. Rossman. "Extensions to
EPANET for Fate and Transport of Multiple Interact-
ing Chemical or Biological Components." Proceed-
ings, ASCE-EWRI World Water & Environmental
Resources Congress, Salt Lake City, UT. 2004.
Van Bloemen Waanders, B., G. Hammond, J. Shadid,
S. Collis, and R. Murray. "A Comparison of Navier-
Stokes and Network Models to Predict Chemical
Transport in Municipal Water Distribution Systems."
Proceedings, ASCE-EWRI World Water & Environ-
mental Resources Congress, Anchorage, AK. 2005.
Vasconcelos, J.J., L.A. Rossman, WM. Grayman, P.P.
Boulos, and R.M. Clark. "Kinetics of Chlorine
Decay," Journal ofAWWA, 89(7):54-65. 1997.
Vose, D. Risk Analysis - A Quantitative Guide. John
Wiley & Sons. Chichester, UK. 2000.
Walski, T.M., D.V. Chase, D.A. Savic, W. Grayman, S.
Beckwith, and E. Koelle. Advanced Water Distribu-
tion Modeling and Management. Haestad Press,
Waterbury, CT. 2003.
Wood, D.J. Computer Analysis of Flow in Pipe
Networks. Department of Civil Engineering, Univer-
sity of Kentucky, Lexington, KY. 1980a.
Wood, D.J. "Slurry Flow in Pipe Networks." Journal
of Hydraulics, ASCE, 106(1). May 1980b.
2-21
-------
A Reference Guide for Utilities
Chapter 3
Tracer Studies for Distribution System Evaluation
Tracers have been used for decades to determine
flow, travel time, and dispersion in surface waters
and groundwater. Tracers can be of various types,
ranging from a physical object that can be visually
detected in a stream or river to dyes or other
chemicals whose concentrations can be monitored
using special instrumentation. Fluorescent dyes
have been used for many years to measure velocity
and tidal movement in streams and estuaries. Use
of tracers to understand the hydraulic movement in
drinking water treatment unit processes or distribu-
tion systems is a more recent development. When
tracers are used in drinking water, care must be
taken to ensure that they will have no adverse
health effects and that their use does not result in
any violations of primary and/or secondary drink-
ing water MCLs.
Tracers have been used in drinking water to
estimate the travel time through various water
treatment unit processes including clearwells
(Teefy and Singer, 1990; Teefy, 1996; DiGiano et
al., 2005). Tracer studies have also been conducted
in distribution system tanks and reservoirs in an
attempt to understand their mixing characteristics
(Grayman et al., 1996; Boulos et al., 1996). They
have also been used in water distribution networks
to provide insight into the complex movement of
water in a distribution system, to determine travel
times, and to assist in calibration of distribution
system hydraulic models (Clark et al., 1993;
DeGiano et al., 2005; Vasconcelos et al., 1997;
Grayman, 2001). For example, Boccelli et al.
(2004) and Sautner et al. (2005) have used dual
tracers injected into water distribution systems to
assess travel time and characterize flow patterns in
support of epidemiological investigations. With the
recent interest in homeland security issues, tracers
are being used to simulate the movement and
impacts of accidental or intentional contamination
of water distribution systems (Panguluri et al.,
2005).
Conducting a distribution system tracer study
involves (1) injecting the tracer into a pipe up-
stream of the area to be studied, (2) shutting off or
reducing a continuous chemical feed at the water
treatment plant, or (3) use of a naturally occurring
substance in source water. The concentration is
measured over time at various locations in the
water distribution network as it moves through the
study area. To be successful, a tracer study requires
careful planning and implementation. This chapter
provides information and guidance on planning
and conducting tracer studies in drinking water
distribution systems.
Tracer studies in distribution systems may provide a
wide variety of useful information, including the
following:
• Calculating travel time, residence time, or water
age in a network.
• Calibrating a hydraulic model.
• Defining zones in a network served by a
particular source and/or assessing the degree of
blending with water from other sources.
• Determining the impacts of accidental or
intentional contamination.
• Identifying appropriate sampling locations
within the water distribution network.
Tracer studies may also assist water utilities in
complying with various regulatory requirements. For
example, the DBPR2 IDSE draft Guidance Manual
(EPA, 2003a) recognizes the use of tracers as a means
of calibrating models and predicting residence time as
a partial substitute for required field monitoring.
Several rules and regulations (both existing and
proposed) are currently being reviewed, such as the
TCR and a proposed distribution system rule. Water
quality modeling and model calibration are likely to
play a role in the development and/or promulgation
of these rules.
The scope of a tracer study may vary considerably
depending upon the study needs, size, and complex-
ity of the distribution network being evaluated. A
study area may consist of a single stretch of pipe, an
entire neighborhood, a portion of a large distribution
system, a pressure zone, or in some cases, the entire
distribution network. The resources required to
conduct a tracer study will vary with the extent,
complexity of the study, and the test equipment used.
Careful planning and implementation are critical in
all cases to ensure meaningful results. Section 3.1 of
this chapter contains information that can be used
during the planning phases of a tracer study. Section
3.2 provides a summary of the tasks associated with
executing a tracer study. Section 3.3 presents typical
costs associated with conducting a tracer study.
Finally, Section 3.4 presents a summary, conclusions,
and recommendations for this chapter. The use of
tracer study data for model calibration/validation is
described in Chapter 4.
3-1
-------
A Reference Guide for Utilities
3.1 Planning and Designing a
Distribution System Tracer
Study
The initial step in any tracer study is a planning and
design phase during which study-specific logistical
details are identified and addressed. These details
should be presented in a comprehensive manner in a
planning document or work plan that can be reviewed
and commented on by parties that may have an
interest in the tracer study (e.g., team members, water
utility staff and managers, and state regulatory
officials). Planning and design-phase elements may
include the following:
• Establishing study objectives and timeline.
• Forming a study team.
• Defining study area characteristics.
• Selecting tracer material.
• Selecting field equipment and procedures.
• Developing a detailed study design.
• Addressing agency and public notification.
The details of each of these tasks are described in the
following sub-sections.
3.1.1 Establishing Study Objectives and Time-
Line
A clear statement of the study objectives should be
developed, even before logistical planning begins.
For example, an objective statement might read
"determine travel times from the Lincoln Water
Treatment Plant to key locations (transmission mains
and representative local mains) in the Washington
Pressure Zone under typical summer operation." Such
a statement provides a clear understanding of the
study's overall goals and objectives. A study objec-
tive may also be more specific and define additional
key elements such as tracer material, dosage, and
injection duration.
Depending upon the objective, an approximate time-
line (schedule) for the study should be formulated.
Frequently, external constraints such as weather,
system operation, and availability of personnel/
equipment may influence this timeline. In other
cases, the project timeline may depend upon the
specific objective of the study. For example, if the
maximum community exposure to a contamination
event is being studied, the timeline should be
consistent with the season and time during which the
event is likely to occur. If the study is intended to
identify locations in the system where the lowest
chlorine residuals are found, the study should be
conducted during a period when minimum chlorine
residuals occur. However, it is not always possible to
conduct a tracer study to match system conditions
that coincide with the study time-frame. Therefore, a
reasonable alternative is to use the tracer to calibrate a
study-area-specific network model, under a given set
of conditions, that can be used to simulate other
critical events under different conditions.
In mid-western U. S., October-November is the best
time-frame to conduct a tracer study in a residential
area. During this time, the utility has greater operational
flexibility because it is not stressed by high demands,
weather is conducive to outdoor activity, and cold
weather pipe breaks are minimal.
3.1.2 Forming a Study Team
A "tracer study team" should be formed at the
beginning of the project. Depending on the size and
scope of the study, the size of the team may vary from
as few as three members to a sizable group of as many
as twenty members. However, the range of functions
and responsibilities that must be considered are
approximately the same in all types of studies. The
team makeup must include members with expertise for
planning and carrying out the following activities
and functions: understanding study area distribution
system and treatment operations; conducting prelimi-
nary modeling studies; selecting, acquiring, and
installing field equipment; managing and organizing
field crews; performing field sampling; conducting
laboratory analysis; analyzing and reporting results;
and performing communications and notifications.
Study teams may be made up of water utility person-
nel, consulting engineering firm personnel, contractor
staff, students from universities, and in some cases,
federal or state governmental agency employees.
Specific responsibilities and roles should be assigned
to each team member. It is recommended that the
study team meet on a regular basis to ensure that the
task deadlines are met and the study objectives are
achievable. If the tracer study includes new or never-
before used equipment, training sessions for study
team members should be included as part of study
timeline and activities.
3.1.3 Defining Study Area Characteristics
After the study team is formed, perhaps the first task
to be undertaken is to identify the key characteristics
of the study area. These characteristics include: the
piping system network, pumping and storage opera-
tions, inflow and outflow through study area bound-
aries, temporal and spatial variations in water con-
sumption, presence of large water users that may
significantly impact water use patterns, and the
3-2
-------
A Reference Guide for Utilities
When planning a tracer study, the effects of distribu-
tion system tanks and reservoirs should be considered
(Grayman at al., 2004). When a tracer enters a tank in
the inflow, it mixes with the distributed water and then
exits the tank at a different concentration during the
subsequent draw cycles. Mixing in the tank may be
rapid and complete or there may be short-circuiting or
plug flow behavior that affects the concentration in the
effluent. Various mathematical tools such as CFD
models may be applied to estimate the mixing charac-
teristics of a tank and the effects on tracer concentra-
tion during discharge periods (Grayman et al., 2004).
Distribution system models such as EPANET allow the
user to simulate mixing in tanks by several alternative
conceptual and simplified models such as completely
and instantaneously mixed, short circuiting, plug flow,
and multiple compartment mixing. The effects of tanks
can impact the needed tracer dosage rate and injection
duration and the subsequent sampling frequency and
duration in parts of the distribution system impacted
by the tank. During the tracer study, the impacts of
mixing in the tank can be determined by sampling in
the inflow and outflow lines, and in some cases,
internally within the tank.
geography and local features associated with the
study area that could potentially constrain field
activities.
A large commercial user such as a golf course in the
neighborhood may impact the study events.
There are several tools and procedures that can be
applied to improve the team's understanding of the
target water distribution system area prior to conduct-
ing the tracer study. If a hydraulic model of the
distribution system (under study) is available, it
would be very helpful to use the model to simulate
the tracer study under expected conditions. Examina-
tion of documents, such as master plans or operational
reports, can also shed light on how the water system
behaves. The study team or key members of the study
team should also tour the study site with as-built pipe
drawings to identify potential locations for safely
installing field injection equipment, as well as flow
and tracer monitoring equipment.
3.1.4 Selecting Tracer Material
Criteria that can influence the selection of a particular
tracer include:
• regulatory requirements,
• analytical methods and instruments available
for measuring tracer concentration,
• injection and storage requirements,
• chemical reactivity,
• chemical composition of the finished water,
• overall cost, and
• public perception.
Ideally, a tracer should be inexpensive, nonreactive
with both water and distribution system materials,
safe to drink when dissolved in water, easily dispersed
in water, aesthetically acceptable to customers, able to
meet all drinking water regulations, and inexpen-
sively and accurately monitored in the field by
manual and automated methods. There is no one
tracer that will meet all of these criteria for a given
study. Frequently, there are tradeoffs among the
criteria listed above that must be assessed when
selecting a tracer. The tracer to be used in the study
should be determined early in the planning stage, and
approval for its use received from the water utility and
state regulatory agencies.
Tracers may fall into three broad categories: a
chemical that is normally added to the water during
the treatment process and that may be temporarily
shut off during the study; a chemical that is added to
the water by the team during the study; or a naturally
occurring substance in the source water that may be
adjusted in some manner to create a tracer.
The most commonly used tracers are fluoride, calcium
chloride, and sodium chloride.
3.1.4.1 Fluoride
Fluoride is frequently added to water supplies
because of its health benefits, but can be turned off
for short periods, thereby making the non-fluori-
dated water a tracer in the system. When fluorida-
tion is not practiced, fluoride can be added to the
water system and used as a tracer by injection. It is
especially popular with utilities that routinely add
fluoride as part of the treatment process, because
little effort is required to turn the fluoride off and
on. When the fluoride feed is shut off, a front of
low-fluoride water (or no fluoride if there is no
natural background concentration) becomes the
tracer. A second tracer test (or a continuation of the
initial test) can be performed when the fluoride feed
Fluoride can interact with coagulants that have been
added during treatment and in some circumstances
can interact with pipe walls leading to non-conserva-
tive behavior. Thus, when used in systems that do not
generally fluoridate, a field test should be performed
to determine possible interactions with pipes.
3-3
-------
A Reference Guide for Utilities
is turned back on, thus making it possible to
generate two sets of tracer data in one study.
The MCL for fluoride is 4 mg/L. However, if the
secondary MCL of 2 mg/L is exceeded, customers
must be notified. Background levels of fluoride can
vary significantly and actually exceed the secondary
MCL in some geographic areas.
In cases where a utility is not permitted to completely
shut off the fluoride feed, it may be feasible to
increase the fluoride feed prior to the tracer study and
to reduce the fluoride feed during the test. Care
should be exercised to avoid exceeding the secondary
MCL. However, there must be a sufficient change in
the fluoride concentration feed in order to trace the
change through the system. Thus, for example, a
decrease in feed concentration from 1.2 mg/L to 0.8
mg/L may not be sufficient, but a decrease in concen-
tration from 1.5 mg/L to 0.5 mg/L may be adequate. A
change in fluoride dosage may have to be pre-
approved by state regulators. Depending upon the
duration of the study, the state agency may choose to
allow a temporary shutoff or set a specific lowest
allowable-fluoride-concentration requirement.
In most treatment plants, fluoride is injected prior to a
final clearwell. As a result, when the feed is shut off as
a part of the tracer study, there is both a time delay
and a gradual change in concentration in the clearwell
discharge as the non-fluoridated and fluoridated water
mix. Therefore, wherever and whenever possible, the
clearwell should be operated at minimum water levels
during the tracer test in order to achieve a relatively
sharp front of non-fluoridated water leaving the
In a study conducted in the Cheshire service area of the
South Central Connecticut Regional Water Authority
(SCCRWA) in 1989, the fluoride feed was turned off to
provide a tracer to validate a hydraulic and water
quality model of their water distribution system (Clark
et al., 1991). This study was among the first applica-
tions of water quality models in the world. SCCRWA
normally added fluoride at a level of approximately 1
mg/L. For purposes of the model validation study, the
fluoride feed was turned off for a period of 7 days and
then turned back on with sampling occurring for an
additional 7 days. This approach yielded, in effect, two
tracer fronts. During the study, grab samples were taken
every few hours at 16 hydrants, two well fields, one
tank, four continuous analyzer sites, and daily at 19
"deadend" sites. Additionally, experimental units were
installed at a few sheltered sites to automatically
measure fluoride concentrations and to take discrete
samples for later analysis. A total of 2,150 fluoride grab
samples were taken during the study and analyzed in
the laboratory.
treatment plant. It is also important to evaluate the
impact of travel through finished water storage
reservoirs on the concentration of tracer during the
study. An alternative is to inject fluoride solution
(e.g., sodium fluoride) at a point in the main transmis-
sion line downstream of the clearwell where both flow
and injection rate can be simultaneously monitored
and measured.
Ion-selective electrodes (ISE) can be used in conjunc-
tion with data loggers to provide continuous monitor-
ing capability. At present, however, these instruments
are relatively expensive (approximately $5,000 to
$10,000 each) and have only been used extensively
in large-scale tracer studies (Maslia et al., 2005;
Sautner et al., 2005). Generally, grab samples are
taken and analysis is performed manually in the field
or laboratory.
Under some circumstances, fluoride is not a fully
conservative chemical. In one study (Vasconcelos et
al., 1996) in a system that did not normally fluoridate,
a 13-hour pulse (step input of limited duration) of
fluoride was injected into the feed line of a pressure
zone. Field measurements of fluoride concentrations
in the zone during the study indicated a significant
loss of fluoride. It was postulated that some of the
fluoride was deposited on the pipe wall. In a
followup study, this problem was virtually eliminated
by injecting fluoride over a period of several days
prior to the actual study in order to pre-condition the
pipes.
3.1.4.2 Calcium Chloride
Calcium chloride (CaCl2) has been used in many
tracer studies throughout the U.S. It is considered to
be safe and relatively easy to handle. Generally, a
food grade substance is required. It can be purchased
as a liquid (typically a 30 to 35% solution) or as a
powder that can be mixed with water to form a
solution.
If calcium chloride is chosen as a tracer, the study
personnel should be aware of the secondary drinking
water MCL for chloride (250 mg/L). A target that is
less than the secondary MCL should be set in order to
provide a safety factor. Where chloride levels are
high, calcium chloride may not be an appropriate
choice for a tracer.
Grayman et al., (2000) utilized calcium chloride as a
tracer in two studies of mixing in distribution system
tanks. In both studies, the chemical was injected into
the inflow pipe of the tank during the fill cycles, and
conductivity and chloride were measured at locations
within the tank. Calcium chloride has recently been
used in several distribution system studies (Panguluri
et al., 2005; Maslia et al., 2005; and Sautner et al.,
2005).
3-4
-------
A Reference Guide for Utilities
Calcium chloride can be monitored by measuring
conductivity, or by measuring the calcium or chloride
ion (Standard Methods, 1998). Conductivity is
typically the easiest of these parameters to measure
and is most amenable to inexpensive continuous
monitors. However, conductivity is not a truly linear
parameter (i.e., if a beaker of water of conductivity 100
mS/cm is combined with a like volume of water with a
conductivity of 300 mS/cm, the conductivity of the
resulting solution will not be exactly 200 mS/cm). As
a result, distribution system models (that all assume
linearity) can only approximately represent conductiv-
ity. Therefore, when using conductivity as the
measured parameter, the options are to accept the
linear approximation or convert conductivity to a true
linear parameter such as chloride or calcium. If the
former option is chosen, the amount of resulting error
should be established in laboratory tests of waters of
varying conductivity. If the latter option is chosen,
the relationship between conductivity and chloride (or
calcium) must be established in the laboratory. It
should also be noted that most field devices are set up
to measure specific conductance instead of conductiv-
ity (conductivity is temperature sensitive, whereas
specific conductance is referenced to 25°C). For the
purposes of this document, conductivity is assumed to
represent specific conductance.
3.1.4.3 Sodium Chloride
Sodium chloride (NaCl) can be used as a tracer and
has many characteristics similar to calcium chloride
in that it can be traced by monitoring for conductivity
or for the concentration of the chloride or sodium ion.
The allowable concentration for sodium chloride is
also limited by the secondary MCL for chloride and
the potential health impacts of elevated sodium
In a recent tracer study in Hillsborough County,
Florida, two separate tracer chemicals were used to
study the movement of water in a large distribution
system (Boccelli et al., 2004). Approximately 2,200
gallons of a saturated NaCl solution was injected into
the finished water of a treatment plant as a series of
four pulses ranging in duration from 1 to 3 hours over
a 24-hour period. Simultaneously, the normal fluoride
feed was shut off at the plant. Continuous conductiv-
ity monitors were installed at 14 locations in the
distribution system to monitor for the NaCl tracer.
Grab samples were taken to monitor the low fluoride
front as it moved through the system and to evaluate
water quality changes. The resulting extensive
hydraulic and water quality database is being used to
calibrate a hydraulic and water quality model of the
system (Boccelli and Uber, 2005).
levels. EPA reports that taste thresholds for sodium
vary significantly among individuals, ranging from
30 to 460 mg/L (EPA, 2003b).
3.1.4.4 Other Chemicals That May be Added as
Tracers
Other chemicals added as part of a tracer study
include lithium chloride and chlorine. Lithium
chloride is a popular tracer in the United Kingdom but
is used less frequently in the U.S., partly because of
the public perception of lithium as a medical pharma-
ceutical. There are no field techniques for measuring
lithium, and it is not easily amenable to automated
continuous measurement. Samples must be collected
and lithium concentrations measured in the laboratory.
Chlorine is commonly used as a disinfectant in many
water systems. Because chlorine is reactive, it will
decay over time. Under some circumstances, however,
it can be used effectively as a tracer. It is most
effective in a water where chlorine is not highly
reactive (low decay rate) with either the water or
distribution system material, and where the concentra-
tion levels can be increased above the normal level to
create a front of water with a high chlorine concentra-
tion propagating through the system. However, in no
case should the chlorine or chloramine be decreased
to a level that may affect the disinfection process
(Ferguson and DiGiano, 2005). Again, any tracer
study should first be approved by the state regulators.
3.1.4.5 Naturally or Normally Occurring Tracers
Perhaps the most difficult part of conducting a tracer
study is obtaining permission to add a chemical and
then injecting the tracer into the system at a concen-
tration consistent with regulations. Much of this
effort can be avoided if there is a natural tracer
available. Natural tracers are generally site-specific,
but many options do exist and should be explored.
The most common situation is the existence of
multiple sources of water with different chemical
signatures or if a change is planned in the chemical
signature at a single source. Examples of these
situations are described below.
Some of the chemical signatures that may be used to
differentiate between sources include THM concen-
trations, hardness, conductivity, and treatment
coagulant. Sampling in the distribution system for
these "tracers" will provide information on zones
served by each of the sources and the extent and
variation of the mixing that takes place in these zones
over time. Alternatively, if one water source can be
turned off for a period of time until the other source
has reached chemical equilibrium throughout the
system, the original source can be turned back on and
used as a tracer as it propagates through the system.
One of the first uses of natural tracers was in the North
3-5
-------
A Reference Guide for Utilities
Making a major change in the incoming water supply
such as a change in source water or modifying treatment
may provide an opportunity to conduct a tracer test.
The increased use of chloramines as a secondary
disinfectant, to reduce the formation of DBFs, intro-
duces another potential tracer opportunity. When a
water utility switches from chlorine to chloramines (or
vice versa), the chemical signature of the water changes
and can be monitored by measuring both free and total
chlorine. Namely, with chloramination, total chlorine is
typically much higher than free chlorine, while with free
chlorination, free and total chlorine will typically be
very similar. A tracer study can be conducted when a
system first adopts chloramination. Alternatively, many
water utilities routinely switch back from
chloramination to chlorine (e.g., annually for a month)
in order to kill ammonia-oxidizing bacteria and thus
reduce the chances of nitrification. This provides a
recurring opportunity to conduct such a tracer study.
Penn Water Authority (NPWA) located in Lansdale, PA
(Clark and Coyle, 1990). A field research project was
conducted by EPA and NPWA that resulted in the
development of a series of models that were used to
study contaminant propagation in the water distribu-
tion system. The utility used a combination of
groundwater with high levels of hardness and surface
water containing higher levels of THMs. This resulted
in two sources of water with very different quality
characteristics. By monitoring changes in water
quality that occurred at selected sampling points in
the utility network, it was possible to use hardness
and THM concentrations as tracers to validate the
model.
Another case occurred in the North Marin Water
District (NMWD) in northern California (Clark et al.,
1994) where natural differences in water characteris-
tics were used to serve as a tracer for validation of a
water distribution system model. In this EPA-
sponsored study, the utility used two sources of water
with dramatically different water quality characteris-
tics. The first source, Stafford Lake, has a very high
humic content and thus has a very high THM forma-
tion potential. The other source is the North Marin
Aqueduct with a very low humic content and thus a
very low THM formation potential. The model was
further validated by predicting chlorine residual
losses at various points in the network. In a follow-up
study supported by AwwaRF (Vasconcelos et al.,
1997), the investigators used sodium as a tracer to
validate the model.
DiGiano and Carter (2001) and DiGiano et al. (2005)
traced the flow from two separate treatment plant
sources at the same time by simultaneously reducing
the fluoride feed at one plant while changing the
coagulant added at the other plant. Normally, ferric
chloride (FeCl3) was used as a coagulant at both
plants. During the tracer study, the coagulant at one
plant was changed to aluminum sulfate [A12(S04)J.
Fluoride, sulfate, and chloride were measured
throughout the distribution system.
Water utilities should carefully examine their particu-
lar system to determine if a natural tracer is available
or if source-chemical signatures may be modified to
be used as a tracer.
Sweetwater Authority in southern California took
advantage of a normal changeover in source water
quality to perform a tracer study in their distribution
system (Hatcher et al., 2004). In this case, the utility
semi-annually changes the primary source of their water
supply from local Sweetwater Reservoir raw water to
water provided by the California Aqueduct. These two
sources have very different chemical characteristics;
most significantly, the organic carbon content (i.e.,
humic and fulvic acids) of Sweetwater Lake water is
much higher compared to the raw aqueduct water. The
measurement of molecular organic carbon absorbance at
254 nanometers, utilizing an ultra-violet-visible (UV-
VIS) spectrophotometer, is a surrogate measurement for
the organic carbon content in water. UV-254 measure-
ments were taken from grab samples at the treatment
plant and at 28 sites within the distribution system over
the five-day changeover period. The distribution
system sites included most of the TCR sampling sites in
addition to selected tanks. The resulting database was
used to assess the movement of water in the system, the
travel time throughout the system, boundary zones in
the distribution system between areas served by the
surface water plant and secondary sources, and calibra-
tion/validation of the distribution system model.
3.1.4.6 Comparison of Tracers
Teefy (1996) investigated tracer alternatives for use in
studies of residence time in clearwells and described
the chemical characteristics of the individual tracers.
Table 3-1 summarizes various chemical characteristics
identified in that report.
There are advantages and disadvantages associated
with each of the general types of tracers: conservative
(non reactive) tracers, reactive tracers, chemicals that
are normally added to the water but can be turned off,
and natural chemical signatures in the finished water.
Conservative tracers are more easily modeled than
non-conservative tracers. Natural tracers or chemicals
that can be turned off are easier to use than injected
chemicals. Certain chemicals are more amenable to
continuous monitors. These and other factors should
all be considered when selecting a tracer for a study.
3-6
-------
A Reference Guide for Utilities
Table 3-1. Tracer Characteristics (adapted from Teefy, 1996)
Commonly
available forms
Analytical
methods
Typical
chemical cost
Typical
analytical cost
per sample
Typical
background
levels in water
distribution
systems
Regulatory
limits
Fluoride
H2SiF6
NaF
Na2SiF6
1C,
ISE,
SPADNS method
Food-grade
H2SiF6
$7.6/1 00 Ib-
23.97% liquid1
$140/55 gallons
- 49% liquid2
$188(IC)
$1610(IC)
$1211 (ISE)
$2512 (1C)
0-4 mg/L
4 mg/L SDWA
MCL, 2 mg/L
secondary MCL
Calcium
CaCI2
AA,
1C,
ICP,
EDTA titration
Conductivity
Food-grade
CaCI2
$150/55 gallons
- 35% liquid3
$108(ICP)
$1210(ICPMS)
$5" (ICP)
Varies greatly (1-
300 mg/L), use
only when low
None known.
See limits for
chloride.
Sodium
NaCI
AA,
1C,
ICP,
FEP
Conductivity
Food-grade NaCI
$12/50lb4
$6/50 Ib5
$108(ICP)
$1210(ICPMS)
$5" (ICP)
Varies greatly (1-
500 mg/L)
20 mg/L for
restricted diet
(EPA
recommendation)
Lithium
dry LiCI
AA,
1C,
ICP,
FEP
Lab-grade LiCI6
$22 - $48/500g7
$1 2s (ICP9)
$1210(ICPMS)
$6" (AA13)
Usually below 5
mg/L
None known.
See limits for
chloride.
Chloride
CaCI2
NaCI
KCI
1C,
ISE,
AgN03 titration.
Hg(N03)2
titration
Food-grade NaCI
$12/50lb4
$6/50 Ib5
$188(IC)
$1610(IC)
$1210 (EPA 325.3)
$12" (1C)
Varies greatly (1-
250 mg/L)
250 mg/L
secondary
standard
1 Provided by Lucier Chemical Industries (LCI), Ltd., http://
www.lciltd.com
2 Provided by Bonded Chemicals, Inc., http://www.chemgroup.com/
bci.htm
3 Provided by Benbow Chemical Packaging, Inc., http://
www.benbowchemical.com
4 Provided by Skidmore Sales and Distributing Company, Inc., http:/
/www.skidmore-sales.com
5 Provided by Ulrich Chemical, Inc., http://www.ulrichchem.com
6 Food grade LiCI is not available.
7 Provided by Science Kit & Boreal Laboratories, http://
www.sciencekit.com
8 Provided by Severn Trent Laboratories (STL) North Canton, Ohio,
http://www.stl-inc.com. Prices are based on a large sample volume
(> 500 samples).
9 STL North Canton Laboratory is not certified for Lithium test in
Ohio.
10 Provided by SPL Laboratories, Inc., http://www.spl-inc.com
Prices are based on a large sample volume (> 500 samples).
11 Provided by Environmental Enterprises, Inc., http://
www.eeienv.com Prices are based on a large sample volume (>
500 samples).
12 Provided by FOH Environmental Laboratory for the CDC study at
Camp Lejeune, NC. http://www.foh.dhhs.gov/. The analytical cost
per sample includes cost for providing a sample bottle and report.
13 Environmental Enterprises, Inc. is not certified for Lithium test.
Note: Tracers
CaCl2
H2SiF6
KCI
LiCI
NaF
Na2SiF6
NaCI
Analytical
AA
AgN03
EDTA
FEP
1C
ICP
ISE
SPADNS
calcium chloride
hydrofluosilicic acid
potassium chloride
lithium chloride
sodium fluoride
sodium silicofluoride
sodium chloride
methods
atomic absorption spectrometry
silver nitrate
ethylenediaminetetraacetic acid
flame emission photometric method
Hg(N03)2 mercuric nitrate
ion chromatography
inductively coupled plasma
ion selective electrode
Trisodium (4,5-Dihydroxy-3-[(p-
sulfophenyl)-2,7-) naphthalene
disulfonic acid
3-7
-------
A Reference Guide for Utilities
After investigating tracer options and selecting the
most appropriate tracer, the governing state drinking
water agency should be contacted. The agency
should be provided with the specifics regarding the
proposed study including location(s), proposed time-
line^) and selected tracer material. Once agreement
has been reached and consent is received, the study
team can then proceed with the next steps in the
planning process.
3.1.5 Selecting Field Equipment and Procedures
Once a tracer has been selected and approval has been
received from the appropriate water utility managers
and regulatory agencies, specialized equipment must
be identified and procured, including injection
pumps, temporary tracer storage tanks, and various
flow and tracer monitoring equipment (e.g., tracer
chemical, reagents, and/or sample bottles). Vendors
should be contacted for technical information,
equipment availability, and cost quotations for the
required field equipment and analytical instrumenta-
tion. The major decisions to be made and the items to
be purchased prior to the execution of the study are
discussed in the following subsections.
3.1.5.1 Injection Pump(s)
Pumps that are typically used in drinking water
applications can be broadly classified as centrifugal
pumps or positive displacement pumps. The centrifu-
gal pumps produce a head and a flow by increasing
the velocity of the liquid with the help of a rotating
vane impeller. The positive displacement pumps
operate by alternating between filling a cavity and
displacing the volume of liquid in the cavity. The
positive displacement pumps deliver a constant
volume of liquid (for a given speed) against varying
discharge pressure or head. By design, the positive
displacement pumps are better suited to serve as an
injection pump for a tracer study. Examples of
positive displacement pumps include: rotary lobe,
progressing cavity, rotary gear, piston, diaphragm,
screw, and chemical metering pumps (e.g., bellows,
diaphragm, piston, and traveling cylinder).
Selection of the most appropriate positive displace-
ment pump depends upon the injection rate, the
pressure in the receiving system, the chemical
characteristics of the tracer, and local experience and
preferences. Two types of positive displacement
pumps have generally been used in tracer studies:
gear pumps and metering pumps. The final selection
depends upon viscosity of the tracer material,
variability of pressure in the main, dosage accuracy
needs, and other local factors. Furthermore, to control
the drive speed (i.e., dosage), these pumps are
equipped with alternating current (AC) or direct
current (DC) motor. If a pump has an AC motor,
frequency is adjusted; if it is equipped with a DC
motor, voltage is adjusted to control speed.
EPA has used gear pumps equipped with variable
frequency drives in the past with success for conduct-
ing tracer studies. Other studies have reported success
with metering pumps with variable speed or variable
stroke controllers. The pump should be sized in
accordance with the anticipated tracer dosage (for
more details, see Tracer Dosage and Injection Dura-
tion Section 3.2.3) and pressure range in the main
pipe for the selected injection location(s) in the study
area. Depending upon the location and dosage
requirements, more than one size of pump may be
needed (excluding backup pumps).
3.1.5.2 Tracer Storage and Dosage Rate
Measurement
Tracers are available in dry or liquid form. If pur-
chased as a powder, provisions for mixing the powder
with water must be made. If the tracer is purchased in
liquid form, it typically comes in either 55-gallon
drums or in larger containers such as a 330-gallon
tote. If only a small amount of tracer is needed, a
single 55-gallon drum will typically suffice. For
greater accuracy, it is recommended that the tracer be
transferred from 55-gallon drums to a suitably sized
day tank with a sight glass (used to periodically
monitor the total tracer volume dosed). It is easiest to
pump the tracer from a single container rather than
having to switch the pump from container to con-
tainer during the injection process. Details on tracer
dosage calculations are presented in Section 3.2.3.
If a metering pump is purchased, care must be taken so
that the pump flow rate is calibrated for the specific
tracer solution (by the vendor). Furthermore, the
During a tracer study when a tracer chemical is being
injected into the system, in order to meet water quality
regulations and to simplify the modeling, it may be
desirable to maintain a constant tracer concentration in
the receiving pipe. This can be accomplished by
monitoring the resulting concentration in the receiving
pipe and manually adjusting the tracer injection rate or
through the use of a closed-loop system for automati-
cally controlling the injection rate based on flow in the
receiving pipe. The automated process is most
effective at a location where the flow in the pipe is
varying relatively slowly and where a flow meter
exists. A typical situation is the use of an existing
venturi meter that generates a 4-20 milli-ampere (ma)
signal. This signal can be used as input to a controller
that has been calibrated and programmed to control the
stroke or speed of a variable stroke or speed injection
pump. If the flow in the receiving pipe is varying
rapidly over a large flow range, it is difficult for the
closed-loop system to respond quickly.
3-8
-------
A Reference Guide for Utilities
variable area flow meters (rotameters - with floats
contained in an upright conical tube) are relatively
inaccurate for measuring tracer dosage even after
adjustments are made for density and viscosity.
Figure 3-1 shows a "flow tube" that can easily be
custom fabricated and calibrated to accurately
measure the rate of tracer injection. It is recom-
mended that the supply tank also be marked to keep
track of the tracer fluid level. Times should be noted
at each mark so that it is possible to create a mass
balance for the tracer injected during the study.
5 gal —
from
storage
tank
to
injection
pump inlet
Figure 3-1. Flow Calibration Tube.
3.1.5.3 Distribution System Flow Rate
Measurement
In order to calculate the concentration of the tracer in
the receiving pipe, it is necessary to know the flow
rate in the pipe, the injection rate of the tracer, the
injected concentration of the tracer, and the back-
ground concentration in the water before tracer is
added. Flow rate should be measured continuously,
because variations in pipe flow rate can affect tracer
concentration. These fluctuations in flow can be
accommodated by manually adjusting the tracer
injection rate in the field or through the use of a flow-
paced injection pump that responds to the flow in the
receiving pipe.
Placement of additional flow meters or other flow
measuring devices at various points in the system is
recommended. This information will be very useful
during the post-tracer modeling studies and is
invaluable in calibrating a network hydraulic model.
If the existing system does not have an adequate
number of flow meters for purposes of a tracer study,
installation of additional meters is recommended.
Various types of flow meters may be used to
measure flow in pipes. They are categorized as either
non-intrusive or intrusive meters. Portable ultrasonic
flow meters are non-intrusive and provide reasonably
accurate data if the pipe material is conductive and
relatively non-tuberculated. The ultrasonic flow
meter requires suitable upstream/downstream straight
runs of pipe. Insertion flow meters are also an option
for measuring pipe flow rates. Insertion meters are
intrusive, and may be magnetic (magmeters) that are
flange coupled to the pipe or have propellers that
must be inserted through a hole in the pipe. All meters
require that the receiving main pipe be exposed (via
excavation) or that an existing vault be used. If the
injection location is in the vicinity of a reservoir/tank
and the water level changes are available in real time,
it may, in some instances, serve as a rough surrogate
for in-pipe flow measurement. The selected method of
flow measurement must be field tested.
Depending upon the size of the reservoir/tank and the
local demand, the reservoir level changes may not be
fast or accurate and precise enough to determine the
flow rate in real time.
3.1.5.4 Field Measurement of Tracer
Concentration
Tracer concentration may be measured in the field
using either automated monitors that analyze a
sample at a preset frequency, by collecting "grab"
samples, or a combination of both. Grab samples can
be manually analyzed in the field or in the laboratory.
If grab sampling is used during a tracer study, the
sampling team will generally traverse a circuit of
several sampling locations. Using such an approach
will generally yield a sampling frequency of one
sample per station every one to three hours for an
average-sized residential neighborhood (unless
multiple crews are used). Some of the factors that will
influence sampling frequency include the speed at
which the tracer is moving within the distribution
system, the number of sampling crews participating in
the study, the number of sampling sites selected, the
time of the day, and the distance between sampling
sites. Equipment requirements for grab sampling are
minimal and may include the following: coolers, ice,
labeled sample bottles, log books, and temperature
blanks. If samples are to be analyzed in the field, the
sampling teams will need the appropriate analytical
equipment. If samples are to be analyzed in the
laboratory, the team will need the means to properly
store and transport samples to a central laboratory.
The quality assurance project plan (QAPP) may
require duplicate or split samples for some or all of
the primary samples. When taking a grab sample, care
must be taken to flush the tap for a sufficient time to
ensure that the sample is representative of the
distribution main rather than the service lines.
3-9
-------
A Reference Guide for Utilities
Reliance solely on grab sampling may be impractical
if the study area is large, the tracer front is moving
rapidly, or a high frequency of sampling is desired. In
these cases, continuous automated monitoring may be
the best choice although some grab samples for
quality assurance and quality control are recom-
mended. If calcium chloride or sodium chloride is the
tracer selected, an online specific-conductivity meter
equipped with an associated data logger is recom-
mended. Automated monitors are available if
chlorine residual is used as a tracer. There are also
automated monitors available if fluoride is used as a
tracer, but there has been relatively limited use under
field conditions. Since most automated monitors
require a continuous side stream (rather than being
inserted directly into a main), the drainage flow from
the monitor must be discharged into a sewer, into the
street and subsequently into a storm drain, or into a
pervious area. This discharge can be an added
complication during cold weather when it may freeze.
Since this discharge stream is generally chlorinated or
chloraminated, regulations may control discharge into
natural water courses. Additionally, this discharge
flow may have to be accounted for if the data set is
being used to calibrate a distribution system model,
and the quantity of discharge through a particular
meter is significant relative to the demand in the
vicinity of the meter. If the total drainage discharge is
significant for the purposes of modeling, provisions
for continuously or manually measuring the amount
of flow being bypassed are needed.
Potential grab and online sampling sites include:
dedicated sampling taps, hydrants, pump stations,
tank inlet-outlet lines, and faucets located inside or
outside of buildings. Figure 3-2 depicts an automated
monitoring station used by EPA. This figure illus-
trates the case where the sampling tap is allowed to
to drain
data logger
data
display
conductivity
Probe
Figure 3-2. Automated Monitoring Station.
EPA and GCWW have pioneered the use of online
monitors as a central focus for distribution system
tracer studies. In a series of field tests, EPA and GCWW
injected calcium chloride tracer into the water system
and followed the movement of the tracer using auto-
mated conductivity meters strategically placed
throughout the study area. Three separate studies were
conducted in a large water system representing a small
highly urbanized area, a small dead-end suburban area,
and a large suburban pressure zone. Based on the
success of these studies, similar tracer studies have
been conducted utilizing a combination of online
monitors and grab samples by the CDC using both
fluoride and sodium chloride as tracers in Hillsborough
County, Florida (Boccelli et al., 2004) and by the Agency
for Toxic Substances and Disease Registry (ATSDR) using
fluoride and calcium chloride at a large military base in
North Carolina (Maslia et al., 2005; Sautner et al., 2005).
run continuously throughout the study with the water
going to a drain. The flow rate to or through the
sampling tap must be sufficient to minimize the travel
time from the main to the monitor.
Online, automated sampling programs should be
complemented with a grab sampling program to add a
degree of confidence in measured data and to supple-
ment field data at additional locations or at the
automated monitor stations if they fail to record
correctly.
3.1.6 Developing a Detailed Study Design
A key element in planning and designing a tracer
study is the preparation of a study design document.
This document serves as the overall plan for conduct-
ing a tracer study and thus, the roadmap for execution
of the study. Three important study-specific parts of
the design plan that may be required before the
execution phase are a QAPP, a Health and Safety
Project Plan (HSPP), and a contingency plan. The
contingency plan describes the actions to be taken if
unexpected events occur; for example, if distribution
system concentrations of the tracer exceed the MCL
for chloride or fluoride. The HSPP should at a
minimum define the job hazards that might be
encountered and the controls, protective equipment,
sample handling and work practices, safety review
procedures, and emergency procedures to be em-
ployed during the study.
The QAPP should clearly define the project objectives,
organization, experimental approach, sampling
procedures, analytical methods, protocols, instrument
calibration requirements, data reporting, data reduction,
and data verification procedures.
3-10
-------
A Reference Guide for Utilities
3.1.7 Addressing Agency and Public
Notification
Appropriate agencies, including fire and police
departments, should be notified prior to the com-
mencement of field activities. With heightened
awareness of security, all people participating should
have a valid identification and contact information. A
standard statement concerning the study should be
developed and provided to all team members in case
they receive inquires at the study site. This same
statement should be used by utility personnel to
answer any telephone inquires that might be received.
. summary information card may be provided to the
study participants that could be handed out to the
public during the study (if requested). This minimizes
the risks of mis-communication.
If the injection site or installation of meters requires
excavation, the study team must obtain the necessary
permits and approvals. This is especially important if
any of the sites are in a residential neighborhood or
near a busy street or road. Care should be taken in all
cases to provide adequate traffic control. Safety is of
paramount consideration.
3.2 Executing a Tracer Study
The team should first become familiar with the
detailed study design documents discussed in Section
3.1.6. Based on these documents, there are several
tasks that need to be completed during the execution
phase of a tracer study. These tasks include:
• Procurement, setup, testing, and disinfection of
study equipment (including pumps, storage
tanks, chemicals, reagents, tubing, connectors,
and continuous tracer monitoring stations).
• Installation of field equipment and testing (both
flow and tracer monitoring equipment to
confirm study-specific distribution system
operation and flow stability).
• Tracer dosage and injection duration
calculations.
• "Dry runs" and planned tracer injection events.
• Real-time field assessments, sampling, and
analysis.
• Equipment demobilization, initiation of data
collection, reduction, and verification process.
These specific execution subtasks are further dis-
cussed in the following sub-sections.
A tailgate safety meeting before commencement of any
field work is the best method to increase awareness.
3.2.1 Procurement, Setup, Testing and
Disinfection of Study Equipment
Field equipment identified under Section 3.1.5 and its
subsections should be procured on a timeline such
that the items arrive several weeks before the planned
study date, especially the monitoring and injection
equipment that may require assembly. An early
arrival will ensure that the equipment can be properly
configured and tested before field use.
Unless pre-calibrated flow-paced injection equipment
is purchased (or if the study does not require injection
equipment - as in the case of using naturally/
normally occurring tracers), the study team should
obtain an appropriate injection pump setup. Figure 3-
3 shows a picture of a tracer injection system used by
EPA for field tests. This setup should be calibrated in
the lab to compute the speed-specific dosage rate
using the tracer solution. If appropriate, a flow-
calibration tube should also be fabricated to confirm
the flow in the field. Figure 3-3 also depicts a flow-
tube used by EPA.
Figure 3-3. Tracer Injection Setup (Storage Tank,
Calibration Tube and Feed Pump).
3-11
-------
A Reference Guide for Utilities
Concurrently, if applicable, the team should initiate
the fabrication of the automated tracer monitoring
stations. These stations are typically equipped with a
probe for measuring the tracer (or a surrogate param-
eter such as conductivity), associated data logger, and
batteries (for powering the probe and the data logger).
If accurate measurement of flow through the auto-
mated monitoring station is needed, it should be
augmented with a household-style water meter and
logger. The equipment should be housed in a secure
lock box to protect it during the field study. Figure 3-
2 shows an automated monitoring station used by
EPA and GCWW to conduct a tracer study. The entire
setup should be tested in the lab to ensure proper
operation and battery capacity to maintain uninter-
rupted operation.
The grab sampling, laboratory equipment, tracer
storage tanks, transportation equipment, and
arrangements should be procured and set up. The
field equipment hookup, including interconnec-
tions between the tracer storage tank, injection
pump, and flow-tube, should be leak tested. The
equipment used for injection should be properly
disinfected and tested prior to field deployment to
ensure that no microbiological contamination
results from the field tests.
If ultrasonic flow meters are procured for field
deployment, the equipment should be set up in a lab
environment to confirm the individual component
operation and approximate battery life. The existing
flow and data acquisition systems to be used in the
field study should be sampled for data accuracy and
field communication.
During the lab testing phase of the field equipment,
the entire field (and backup) crew should familiarize
themselves with proper operating procedures for the
equipment they are designated to operate.
One procedure for equipment disinfection is to
prepare approximately 50 gallons of 50 ppm chlorine
disinfectant solution. This solution is then re-
circulated through the injection pump setup for about
15 minutes. Thereafter, continuously flush the
injection pump using de-ionized water for about 15
minutes. Collect a water sample at the end of the
flush cycle and send it for bacteriological analysis
(Coliform and E. coli) to insure that the disinfection
procedure was successful. For the purposes of
sampling, use sterile sample bottles with a de-
chlorinating agent (e.g., sodium thiosulfate). The de-
chlorinating agent is added to remove any residual
chlorine or other halogen that may continue the
disinfection process in the sample and yield incorrect
test results.
3.2.2 Installation of Field Equipment and
Testing
Prior to the commencement of field activity, a brief
"tailgate" health and safety meeting should be
conducted at the beginning of each day to remind the
crew of potential job hazards. Mobilization of field
equipment for excavations (if required - for installing
main flow meters) should be initiated to allow for the
flow monitoring devices to be installed prior to the
scheduled injection event(s). This time lag will vary
according to the needs of the specific study and could
range from several days to several weeks. The early
installation of flow meters will allow the study team
to capture actual field flow data for performing any
revisions to tracer dosage computations and prelimi-
nary hydraulic modeling analysis. The flow meter
installation location should meet the manufacturer's
recommendations for upstream and downstream
straight lengths of undisturbed pipe. The excavations
should be performed in accordance with the HSPP
Appropriate drainage for the excavated pits should be
arranged in case rain is forecast during the study
period.
The measured field flow data should be utilized to
confirm the stability and range of flow at the injection
location and other major branches of the system where
flow is monitored. It may be necessary to operate the
distribution system under specified conditions in
order to achieve optimum results during the study.
The operational changes that may be required
include: scheduled cycling of tank levels, pumps,
and valves. Time required for the deployment of the
automated monitoring stations prior to the start of the
tracer tests is dependent upon several factors, includ-
ing the number of monitoring stations, the distances
between stations, the ease of attaching the stations to
the sampling hydrants, and the effort required to
calibrate the monitoring equipment. If feasible and
consistent with normal operating policies, the system
should be operated to avoid frequent abrupt changes
in flow such as would be associated with a pump that
was cycling on and off very rapidly.
A day or two prior to the execution of the tracer
injection event, the study team should fully deploy
the continuous monitoring stations (if used). These
stations should be hooked up at the designated
sampling locations and data logs should be checked
to ensure data are being collected. Flow through a
monitoring station should be sufficient to minimize
the time delay in detecting the injected tracer
between the main and the sampling location. Experi-
ence has shown that 1 to 2 gallons per minute (gpm) is
usually sufficient. The field crew should also test the
coverage and reliability of field communication
devices (such as cellular phones) in the designated
study area.
3-12
-------
A Reference Guide for Utilities
3.2.3 Tracer Dosage and Injection Duration
Calculations
Factors affecting the amount of tracer required for the
study include the duration of the injection, the flow
rate in the receiving pipe, and the target concentra-
tion in the distributed water. This target concentra-
tion should be consistent with drinking water
standards. For example, if fluoride is being injected
(into a system that does not fluoridate) with a second-
ary MCL of 2 mg/L, a reasonable target concentration
level is 80% of the MCL, i.e., 1.6 mg/L. The injection
rate should be set to meet that goal.
Using the principle of material balance, the resulting
tracer concentration in a receiving pipe downstream
of the point of injection can be calculated as follows:
QD=Qu+QT (Equation 3-1)
Cl U-6*
D =
Where
(Equation 3-2)
QD
QD = flow downstream of injection point, LVT
QJJ = flow upstream of injection point, LVT
QT = flow of tracer solution, LVT
CD = concentration of tracer material downstream
of injection point, M/L3
CB = background concentration of tracer material in
distributed water, M/L3
CT = tracer concentration, M/L3
Equation 3-1 represents continuity and Equation 3-2
represents conservation of mass. As written, these
equations are independent of units for mass (M),
length (L), and time (T) as long as consistent units are
used for computations. However, when tracer
concentrations, injection rates, and injection
duration are used to calculate the required volume of
tracer material purchased, units for flow, concentra-
tion, and time must be commensurate or appropriate
conversion factors must be employed.
For some tracers, the allowable concentration in the
distributed water may be controlled by one of the
dissolved ions that are part of the tracer. For example,
if calcium chloride is the selected tracer, the concen-
tration of the chloride ion in the distributed water
controls the amount of tracer that may be injected.
Injection duration depends upon the size and com-
plexity of the distribution system, and the modeling
objectives of the study. A typical duration can range
from one hour in a small or branched system, to eight
hours or more in a larger, looped system. Some
studies have reported success with a series of pulses.
However, if the duration of the injection is too short
or the series of pulses too close together in time, it is
difficult to separate the tracer fronts as they traverse
different paths at different velocities through the
looped systems. The presence of tanks can also
impact the needed tracer duration since active filling
and drawing can dampen the resulting tracer concen-
tration as it moves through the system.
The injection equipment should be located close to
the main in order to minimize the tracer travel time to
the main. Alternatively, the travel time should be
compensated for during the appropriate phases of the
study evaluation.
3.2.4 Dry Runs and Planned Tracer Injection
Event(s)
Before the planned full-scale tracer injection event is
actually carried out, the project team should consider
conducting a smaller duration dry run injection to
confirm the system operation and expected levels of
tracer concentration. If continuous monitors are to be
used in the study, then during the dry run some or all
of the monitors should be installed and tested. The
timing and duration of the dry run should be such that
the injected pulse should be short and clear the
system well before the actual event is initiated.
The dry run serves as a final systems check and
provides the study team an opportunity to make any
necessary last minute changes prior to the actual
study. Thereafter, the actual full-scale injection event
should be conducted as planned.
3.2.5 Real Time Field Assessments, Sampling,
and Analysis
While the injection event is ongoing, the study team
should carefully monitor the tracer concentration at
the immediate downstream location of the injection to
ensure that there are no significant deviations in the
expected versus observed concentrations in the field.
Field crews should communicate directly with the
system operations. It is critical that the field person-
nel are aware of any changes in system operations that
may affect the study. Unanticipated changes in water
demand may cause the tracer concentration to exceed
target concentration levels. In such an event, the field
crew should be trained to take measures to minimize
any adverse effects. The preventive measures may
include lowering (or stopping) the injection rate, or
achieving appropriate dilution by means of rerouting
water through the distribution system (as appropriate).
Furthermore, any such tracer concentration
exceedances should be confirmed by performing field
grab sample analysis to make sure that the exceedance
is real and not an instrument anomaly. Until the
3-13
-------
A Reference Guide for Utilities
results are confirmed, it is best to err on the safe side
and take preventive measures to maintain water quality.
Periodically, the field crew should take grab samples
and inspect the continuous monitoring stations to
ensure that the equipment is operating properly. The
grab samples should be appropriately handled and
analyzed in the field or transported to the laboratory
for further analysis. The sampling and monitoring
effort should continue well past the conclusion of the
injection event until the tracer is expected (and
observed) to have moved out of the system. This may
take a period of 24 to 48 hours or more after comple-
tion of the injection event.
During the course of the sampling event, it is very
useful to examine and assess the field data on a near
real-time basis. Questions that should be asked
include "Are the results reasonable?" "Is the tracer
moving through the system at a speed consistent with
predictions?" Based on this assessment, modifica-
tions may be made in terms of injection rate, grab
sampling frequency, or study duration.
3.2.6 Equipment De-Mobilization, Initiation of
Data Collection, Reduction, and
Verification Process
After the scheduled injection event(s) are completed,
the field crew should download the data (including
flow and tracer concentrations) from the various
monitoring devices. The data should be spot checked
against field grab sampling data to ensure that there
are no time anomalies or gaps in the data log and the
readings match relatively well.
After the field sampling events are completed, the
crew should de-mobilize the equipment, remove the
automated monitoring stations, refill any excavations,
and restore the system operations to their normal
conditions.
Downloaded data from the field should be processed
according to the QAPP and used for further modeling
and analysis. The use of field data in calibration and
validation of hydraulic and water quality models is
discussed further in Chapter 4.
3.3 Tracer Study Costs
In general, the cost of conducting a tracer study is
proportional to the study area size, number of
monitoring sites, study duration, sophistication and
amount of equipment, and complexity of post-study
analysis. If a study incorporates an injected tracer and
the use of continuous monitors, it can be much more
expensive initially than a study using a natural tracer
and grab samples. However, the injection equipment
and continuous monitoring equipment can be reused
at various locations. These are the cost tradeoffs
between purchase of automated monitoring equip-
ment and labor associated with grab sampling. In
some cases, a larger dataset derived from an auto-
mated monitor is necessary for a detailed analysis.
Cost data presented in this section are intended to
provide the basis for this type of analysis. For the
purposes of this chapter, the overall costs have been
broken down into two distinct categories: equipment
and labor. Material costs are only a fraction of the
total, and therefore, have been combined and in-
cluded with equipment costs for simplicity.
Table 3-2 lists typical equipment and material costs
for those items that may be used in tracer studies. The
unit costs can be easily scaled to the needs of a
specific study. Chemical tracer costs, including
analytical costs, were provided earlier in Table 3-1.
Costs may vary widely among studies. For example,
if it is necessary to purchase or rent a storage tank or a
Table 3-2, Equipment Costs
Equipment & Material
Injection pump
Flow meter (ultrasonic meter for
main pipes)
Excavation, rigging and backfill
(equipment rental per site)
Lab chemicals, batteries and
plumbing supplies (lump sum*)
Automated monitoring box (self
constructed)
Online conductivity ISE, meter
and logger
Automated monitoring station
water flow meter
Online fluoride meter
Safety equipment (e.g., vests,
first aid kits, rain gear, and
flashlights)
Communication equipment (e.g.,
radios and GPS)
Hydrant equipment (e.g.,
wrenches, caps, and hoses)
Transportation (e.g., rental
vehicles)
Tracer storage tanks (depending
upon volume and material)
Unit Cost ($)
$1,000 -$5,000
$7,000 - $9,000
$1,500
$1,000 -$5,000
<$200
$800 -$1,500
$600 - $800
$5,000 -$10,000
$500 -$1,000
$500 -$1,000
$1,000 -$2,000
$500 - $2,000
$500 -$1,000
3-14
-------
A Reference Guide for Utilities
truck, costs will be higher if these types of items are
not readily available. If the study team elects to
analyze samples in-house rather than using an outside
laboratory, the team should balance the cost of labor,
and the cost of additional reagents and chemicals
against the cost of performing the analyses at an
outside commercial laboratory. Labor costs may be
even more variable than equipment and material costs
and are a function of the size and complexity of the
study. In order to provide an easy basis for compari-
son, the labor costs are presented in labor hours (Table
3-3) and include a combination of engineers and
technicians. Labor hours have been estimated for low,
medium, and high-end studies. These estimates are
obtained from actual field studies, as described below.
This approach should allow utilities to make site-
specific cost estimates.
Table 3-3. Representative Labor Hours
for a Range of Studies
Activity
Planning
Setup
Field study
Laboratory
analysis
Post-study
assessment
Total
Low-End
27
-
51
8
24
110
Medium
274
150
604
160
212
1,400
High-End
480
520
370
120
740
2,230
A typical example of a low-end tracer study is
provided by the Sweetwater Authority distribution
system in Southern California (see second sidebar in
Section 3.1.4.5, page 3-6). The Sweetwater system
covers a service area of 28 square miles. The utility
was able to take advantage of a naturally occurring
tracer and used grab samples taken at 28 existing
dedicated sampling sites over a period of 5 days. A
study performed in the 21-square-mile Cheshire
service area of the South Central Connecticut
Regional Water Authority in 1989 (see second sidebar
in Section 3.1.4.1 on page 3-4) provides an example
of a medium-level tracer study. In this case, the
normal fluoride feed was shut off for a period of 7
days (and then turned back on) and grab samples were
taken at intervals of a few hours at 23 sites over a
period of 14 days. An example of a high-end study is
provided by a two-phased field investigation con-
ducted in two suburban areas of GCWW. The first area
is a small (<1 square mile) dead-end system, and the
second area, a 12-square-mile pressure zone. A
calcium chloride tracer was injected and monitored
using a combination of automated conductivity
meters and grab samples. In the smaller area, 20
meters were used and monitoring was conducted over
a 24-hour period. In the second area, 33 meters were
used and two separate tracer injections were con-
ducted over a period of 5 days. Including both
studies, a total of 725 grab samples were taken and
analyzed for conductivity, chloride, and calcium.
Flow was monitored at four locations using ultrasonic
flow meters.
Table 3-3 presents estimated labor hours for these
types of studies. They are divided into the planning
phase (as described in Section 3.1); setup, field work,
and laboratory analysis that together make up the
execution phase (see Section 3.2); and the post-study
modeling, assessment, and report phase. As illus-
trated in this table, there is a significant variation in
the labor hours required to conduct a tracer study. For
example, the low-end labor costs resulted due to the
following study characteristics: naturally occurring
tracer was used, no new equipment was purchased,
existing routine monitoring sites were used, and only
a limited post-study assessment was made. The
medium-sized study included the following character-
istics: a chemical that was routinely added (fluoride)
to the water distribution system was used as the tracer
(by shutting it off), the study required a much longer
period to complete, and since it was the first major
tracer study in the distribution system, it required
significant planning. The high-end study included
the following characteristics: it was the first major
tracer study employing wide-scale use of continuous
monitors; a non-naturally occurring, non-routinely
added chemical was injected as a tracer; and signifi-
cant time was required for acquiring and installing the
equipment. For purposes of this study, a very detailed
post-study data assessment involving processing of
tracer study data, pipe network model calibration and
report preparation required significant labor expendi-
tures. Examples of model calibration efforts associ-
ated with tracer studies are presented in Chapter 4.
3.4 Summary, Conclusions and
Recommendations
Tracers and tracing techniques have been used for
many years in a number of engineering applications
to estimate stream velocity and retention time in
water and water supply unit processes. More recently,
tracers have been used for calibrating drinking water
distribution system hydraulic and water quality
models. For the purposes of this document, it is
assumed that tracer studies are used to calibrate and
validate network models. The calibrated and vali-
dated network models are then used to estimate other
parameters such as water age and travel times.
However, the data from a tracer study can be directly
used to estimate some specific parameters such as
water age (DiGiano et al., 2005). A comprehensive
3-15
-------
A Reference Guide for Utilities
summary of potential uses and regulatory applications
for tracer studies is provided in the first subsection of
this chapter. Drinking water tracers might include
chemicals that are injected into a water distribution
pipe, the temporary shutoff of a chemical additive
currently being added to treated water (such as
fluoride), or significant changes in concentration of
disinfectants, DBFs, or natural compounds. The tracer
methodology selected would significantly impact the
overall costs of the study. Probably, the most expen-
sive option would be to inject a chemical tracer,
monitor it using leased or purchased online instru-
mentation, and conduct the study using contractor
staff. The least expensive approach would be to take
advantage of a natural tracer, monitor the progress of
the tracer by grab sampling, and conduct the study
using primarily in-house staff. Once a tracer "injec-
tion" methodology has been selected, careful plan-
ning and execution will ensure the success of the study.
When planning a tracer study, if the specific steps
outlined in this chapter are followed, they should
greatly increase the potential for a successful study.
These steps include: establishing clear study objec-
tives, forming a study team, defining the study area
characteristics, carefully selecting an appropriate
tracer, selecting the proper field equipment, develop-
ing key planning documents, and ensuring that the
public and affected agencies are notified. Applica-
tion of a distribution and water quality model during
the planning stage is highly recommended to simulate
the approximate behavior that will be expected
during the actual tracer event.
During the execution phase of the study, the follow-
ing issues should be addressed: procurement of
equipment and materials; setup, testing and disinfec-
tion of the procured equipment; availability of
analytical instrumentation and laboratory facilities;
and, finally, the installation, testing, and operation of
field equipment. During the execution phase, it is
important to review and understand how tracer
dosages and injection duration are to be implemented.
Dry runs are highly recommended as a means of
debugging the procedures prior to a full study.
Distribution system tracer studies have been conducted
for over 15 years, but recent technology developments
have improved the efficiency of these studies and
provide promise for greatly expanded applications in
the future. Specific components that will fuel this
expanded use include the following: continuous
monitors that can be easily adapted for use in distribu-
tion systems are being developed and tested, in part in
response to water security concerns; automated meter
reading (AMR) equipment is being installed by many
utilities and could provide more detailed temporal and
spatial consumption data for hydraulic models;
advanced analysis software is evolving that will
facilitate the use of large amounts of continuous data in
calibrating distribution system models; and with
increased availability of these technologies, costs are
expected to decrease so that larger utilities can afford to
purchase and routinely use the equipment, and consult-
ing engineers can affordably offer these services to
smaller utilities.
During the field study, it is important that the study
team be able to assess the progress of the tracer, in real
time, as it propagates through the system. Concise
and consistent communications between tracer study
team members, test coordinator, and water utility staff,
is critical al all times during the test.
In the future, it is highly likely that advances cur-
rently on the horizon will result in significant
increased use of both online tracer (or water quality)
monitors and flow monitoring instrumentation. As
the on-line technology becomes more widely used in
drinking water, the use of network water quality
models will also be more widely accepted. Online
monitoring in conjunction with water quality
modeling will provide an in-depth understanding of
the manner in which water quality changes can be
monitored in a drinking water distribution system.
Also, given the current climate of concern over
distribution water quality from both a regulatory and
security viewpoint, it is reasonable to assume that
there will be increased interest in applying this type of
technology in the water industry.
3-16
-------
A Reference Guide for Utilities
References
Boccelli, D.L., F. Shang, J. G. Uber, A. Orcevic, D.
Moll, S. Hooper, M. Maslia, J. Sautner, B. Blount, and
F. Cardinal!. "Tracer Tests for Network Model
Calibration." Proceedings, ASCE-EWRI Annual
Conference. 2004.
Boccelli, D., and J. Uber. "Incorporating Spatial
Correlation in a Markov Chain Monte Carlo Ap-
proach for Network Model Calibration." Proceedings,
ASCE-EWRI World Water & Environmental Re-
sources Congress, Anchorage, AK. 2005.
Boulos, PR, WM. Grayman, R.W. Bowcock, J.W
Clapp, L.A. Rossman, R.M. Clark, R.A. Deininger, and
A.K. Dhingra. "Hydraulic Mixing and Free Chlorine
Residuals in Reservoirs," Journal of AWWA, 88(7) :48-
59. 1996.
Clark, R.M., WM. Grayman, R. Males, and A. F. Hess.
"Modeling Contaminant Propagation in Drink-
ingWater Distribution Systems," Journal of Environ-
mental Engineering, ASCE, 119(2):349-364. 1993.
Clark, R.M., W.M. Grayman, J.A. Goodrich, R.A.
Deininger, and A.F. Hess. "Field Testing Distribution
of Water Quality Models," Journal of AWWA,
83(7):67-75. 1991.
Clark, R.M., G. Smalley, J.A. Goodrich, R. lull, L.A.
Rossman, J.T. Vasconcelos, and PF. Boulos. "Manag-
ing Water Quality in Distribution Systems: Simulat-
ing TTHM and Chlorine Residual Propagation,"
Journal of Water Supply Research and Technology -
Aqua, 43(4):182-191. 1994.
Clark, R.M., and J.A. Coyle. "Measuring and Model-
ing Variations in Distribution System Water Quality,"
Journal of AWWA, 82(8):46-53. 1990.
DiGiano, FA., and G. Carter. "Tracer Studies to
Measure Water Residence Time in a Distribution
System Supplied by Two Water Treatment Plants."
Proceedings, AWWA Annual Conference. 2001.
DiGiano, F.A., W. Zhang, and A. Travaglia. "Develop-
ment of the mean residence time from tracer studies in
distribution systems." Journal of Water Supply:
Research and Technology - Aqua 54:1-14. 2005.
EPA. The Stage 2 DBPR Initial Distribution System
Evaluation Guidance Manual. 2003a.
EPA. Drinking Water Advisory: Consumer Accept-
ability Advice and Health Effects Analysis on Sodium.
Available at: http://www.epa.gov/safewater/ccl/pdfs/
reg_determinel/support_cc l_sodium_dwreport.pdf.
2003b.
Ferguson, B.A. and F.A. DiGiano. "Impact of tempo-
rary switches from monochloramine to free chlorine
on water quality in distribution systems." Proceed-
ings, AWWA Annual Conference. 2005.
Grayman, W.M., R.A. Deininger, A. Green, PF. Boulos,
R.W. Bowcock, and C.C. Godwin. "Water Quality and
Mixing Models for Tanks and Reservoirs," Journal of
4, 88(7):60-73. 1996.
Grayman, W.M., L.A. Rossman, C. Arnold, R.A.
Deininger, C. Smith, J.F. Smith, and R. Schnipke.
Water Quality Modeling of Distribution System
Storage Facilities. AwwaRF and AWWA. Denver, CO.
2000.
Grayman, W.M. "Use of Tracer Studies and Water
Quality Models to Calibrate a Network Hydraulic
Model," Current Methods, l(l):38-42, Haestad Press,
Waterbury, CT. 2001.
Grayman, W.M., L. A. Rossman, R. A. Deininger, C. D.
Smith, C. N. Arnold, and J. F. Smith. "Mixing and
Aging of Water in Distribution System Storage
Facilities," Journal of AWWA, 96(9):70-80. 2004.
Hatcher, M.D., W. M. Grayman, C. D. Smith, and M.
Mann. "Monitoring and Modeling of the Sweetwater
Authority Distribution System to Assess Water
Quality." Proceedings, AWWA Annual Conference.
2004.
Maslia, M.L., J.B. Sautner, C. Valenzuela, W.M.
Grayman, M.M. Aral, and J.W. Green, Jr. "Use of
Continuous Recording Water-Quality Monitoring
Equipment for Conducting Water-Distribution System
Tracer Tests: The Good, the Bad and the Ugly."
Proceedings, ASCE-EWRI World Water & Environ-
mental Resources Congress, Anchorage, AK. 2005.
Panguluri, S., R. Krishnan, L. Garner, C. Patterson, Y.
Lee, D. Hartman, W. Grayman, R. Clark and H. Piao.
"Using Continuous Monitors for Conducting Tracer
Studies in Water Distribution Systems." Proceedings,
ASCE-EWRI World Water & Environmental Re-
sources Congress, Anchorage, AK. 2005.
Sautner, J.B., M.L. Maslia, C. Valenzuela, W.M.
Grayman, M.M. Aral, and J.W. Green, Jr. "Field
Testing of Water Distribution Systems at U.S. Marine
Corps Base, Camp Lejeune, North Carolina, in
Support of an Epidemiological Study." Proceedings,
ASCE-EWRI World Water & Environmental Re-
sources Congress, Anchorage, AK. 2005.
3-17
-------
A Reference Guide for Utilities
AWWA. Standard Methods for the Examination of
Water and Wastewater, 20^ Edition. Edited by:
Lenore S. Clrsceri, Arnold E. Greenberg, and Andrew
D. Eaton, American Pubic Health Association/AWWA/
Water Environment Federation, pp 2-44 to 2-45.
1998.
Teefy, S.M., and P.C. Singer. "Performance Testing
and Analysis of Tracer Tests to Determine Compliance
of a Disinfection Scheme with the SWTR." Journal of
AWWA, 82(12):88-98. 1990.
Teefy, S.M. Tracer Studies in Water Treatment
Facilities: A Protocol and Case Studies. AwwaRF
and AWWA, Denver, CO. 1996.
Vasconcelos, J.J., L.A. Rossman, WM. Grayman,P.F.
Boulos, and R.M. Clark. "Kinetics of Chlorine
Decay," Journal of AWWA, 89(7):54-65. 1997.
Vasconcelos, J., P. Boulos, W. Grayman, L. Kiene, 0.
Wable, P. Biswas, A. Bahri, L. Rossman, R. Clark, and
J. Goodrich. Characterization and Modeling of
Chlorine Decay in Distribution Systems. AwwaRF.
1996.
3-18
-------
A Reference Guide for Utilities
Chapter 4
Calibration of Distribution System Models
Water distribution system models can be used in a
wide variety of applications to support design,
planning, and analysis tasks. Since these tasks may
result in engineering decisions involving significant
investments, it is important that the model used be an
acceptable representation of the "real world" and that
the modeler have confidence in the model predic-
tions. In order to determine whether a model repre-
sents the real world, it is customary to measure
various system values (e.g., pressure, flow, storage
tank water levels, and chlorine residuals) during field
studies and then compare the field results to model
predictions. If the model adequately predicts the field
measurements under a range of conditions for an
extended period of time, the model is considered to be
calibrated. If there are significant discrepancies
between the measured and modeled data, further
calibration is needed. There are no general standards
for defining what is adequate or what is a significant
discrepancy. However, it is recognized that the level
of calibration required will depend on the use of the
model. A greater degree of calibration is required for
models that are used for detailed analysis, such as
design and water quality predictions, than for models
used for more general planning purposes (e.g., master
planning).
All models are approximations of the actual systems
that are being represented. In a network model, both
the mathematical equations used in the model and the
specific model parameters are only numerical approxi-
mations. For example, the Hazen-Williams equation
used to describe friction headloss is an empirical
relationship that was derived based on laboratory
experiments (Williams and Hazen, 1920). Further-
more, the roughness parameter (C-factor) used in the
Hazen-Williams equation that modelers assign to each
pipe is not known with total certainty because it is
not feasible to examine and test every pipe in the
system. The goal in calibration is to reduce uncer-
tainty in model parameters to a level such that the
accuracy of the model is commensurate with the type
of decisions that will be made based on model
predictions.
The types of model calibration associated with water
distribution system analysis can be categorized in
several ways. The nomenclature depends upon the
adjusted parameters and the technique employed. In
general, calibration can be categorized (or referenced)
as follows:
• Hydraulic and water quality model calibration.
The concept of calibration can be compared to fine
tuning an old fashioned television (TV) set. One knob
on the TV is used for tuning the channel while other
knobs are adjusted to improve color, sharpness, contrast,
and hue. However, in calibrating a network model, there
are far more knobs to adjust as illustrated in Figure 4-1.
ooooooooooooooooo
00000000000000000
ooooooooooooooooo
ooooooooooooooooo
ooooooooooooooooo
ooooooooooooooooo
O Adjustment knobs
• Field data
^—^— Initial model results
^—^— Model results after calibration
Figure 4-1. Conceptual Representation of Calibration.
Some of the knobs may be used to adjust roughness
coefficients for pipes, other knobs to adjust demands
assigned to nodes, while still other knobs may control
valve positions, pump curves, or other parameters that
are not known with complete certainty. Calibrating a
model is an arduous task because there are many knobs
that can be adjusted. Finding the combination of
parameters that results in the best agreement between
measured and modeled results is difficult. This process
is complicated by the fact that there may not be a single
best set of parameters. Extending the TV analogy, the
knobs may be adjusted in order to get the best reception
for one channel. However, when the channel is changed,
the knobs may need to be adjusted to improve the
reception for the new channel. Similarly, with a network
model, a set of parameters may give the best match for
one set of data while other parameters may give better
results for another set of data. Therefore, it is recom-
mended that a modeler first calibrate the model using
one or more sets of field data and then validate it with
an independent set of field data.
4-1
-------
A Reference Guide for Utilities
• Static (steady state) or dynamic (extended
period simulation) calibration.
• Manual or automated calibration.
Hydraulic calibration refers to the process of adjust-
ing the parameters that control the hydraulic behavior
of the model. Similarly, water quality calibration
relates to the process of adjusting parameters used in
the water quality portion of the model. Static or
steady-state calibration relates to calibration of a
model that does not vary over time, or using data that
is collected representing a snapshot in time. Dynamic
or EPS calibration uses time-varying data in the
calibration process. Manual calibration relies upon
the user to investigate the effects of a range of
possible parameter values. Automated calibration
employs optimization techniques to find the set of
parameters that results in the "best" match between
measured and modeled results.
It should be noted that the specific application
method and availability of some of these techniques
will vary depending upon the software used for
modeling and the available network model informa-
tion. Therefore, only the general techniques em-
ployed in each of these types of calibration are
discussed in the following sections. Then, some
example case studies are presented to illustrate their
use. The final section in this chapter discusses future
trends in calibration and the possibility of general
calibration standards.
4.1 Hydraulic and Water Quality
Model Calibration
Hydraulic calibration is essential for any model
simulation to be meaningful. Furthermore, the
distribution system water quality models work in
concert with the hydraulic model and utilize the flow
and velocity information calculated by the hydraulic
model. Thus, if the hydraulic model is not properly
calibrated and results in inaccurate flow and velocity
estimates, the water quality model will not perform
correctly. In fact, water quality modeling is very
sensitive to the underlying hydraulic model. Fre-
quently, a hydraulic model that has been calibrated
sufficiently for applications such as master planning
may require additional calibration before it is
appropriate for use in water quality modeling. The
following subsections describe the parameters and
techniques employed for hydraulic and water quality
model calibration.
4.1.1 Hydraulic Model Calibration
Hydraulic behavior refers to flow conditions in pipes,
valves and pumps, and pressure/head levels at
junctions and tanks. Parameters that are typically set
and adjusted include pipe roughness factors, minor
losses, demands at nodes, the position of isolation
valves (closed or open), control valve settings, pump
curves, and demand patterns. When intially establish-
ing and adjusting these parameters, care should be
taken to keep the values for the parameters within
reasonable bounds. For example, if local experience
shows that the roughness factor for a 20-year old
ductile iron pipe typically falls within a range from
100 to 130, a value that is not within or close to that
range should not be used just to improve the agree-
ment between the measured and modeled data. Use of
unreasonable values may lead to a better match for
one set of data, but will typically not provide a robust
set of parameters that would apply in other situations.
Proper calibration requires that adjustments be made
to the correct parameters. A common mistake occurs
when adjustments are incorrectly made in one set of
parameters in order to match the field results while the
parameters that are actually incorrect are left un-
touched. This process is referred to as "compensating
errors" and should obviously be avoided. Field
verification of suspect parameters (e.g., open or closed
valves) can reduce confusion created by compensat-
ing errors.
An example of compensating errors is an adjustment in
roughness factors in order to compensate for a closed
isolation valve in the system that is represented as open,
or partially open, in the model. In this case, unreason-
ably low values for the Hazen-Williams roughness
coefficients are typically introduced in order to force a
large headloss in the pipes that are actually closed.
Though this may result in approximating the pressure
measurements made in the field, it will introduce other
errors in flow and velocity calculations. Compensating
errors may also result from incorrectly adjusting
demands or other parameters.
4.1.2 Water Quality Model Calibration
Subsequent to the proper calibration of a hydraulic
model, additional calibration of parameters in a water
quality model may be required. The following
parameters are used by water quality models that may
require some degree of calibration:
• Initial Conditions: Defines the water quality
parameter (concentration) at all locations in the
distribution system at the start of the
simulation.
• Reaction Coefficients: Describes how water
quality may vary over time due to chemical,
biological or physical reactions occurring in the
distribution system.
• Source Quality: Defines the water quality
characteristics of the water source over the time
period being simulated.
4-2
-------
A Reference Guide for Utilities
The details of calibration depend upon Table 4-1. Calibration/Input Requirements for Water Quality Models
the type and application of the water
quality model. Calibration require-
ments for each type of modeling are
described below and summarized in
Table 4-1.
• Water age: No explicit water
quality calibration can be
performed because there are no
reaction coefficients. Estimates
of initial water age in tanks and
reservoirs are desirable in order
to shorten the length of the
simulation. Source water age is usually set to
zero for all sources. Water age can be especially
sensitive to inflow-outflow rates for tanks,
mixing characteristics of tanks, and travel times
in dead-end pipes.
When modeling a tank, an important parameter is the
initial age of the water in the tank at the start of the
simulation. This value cannot be measured in the field
but can be estimated by dividing the tank volume by
the volume of water that is exchanged each day.
Frequently, modelers will just assume that the initial
age is zero and run the model for a long period until it
has reached a dynamic equilibrium. This occurs when
the initial water in the tank has been flushed out
entirely through the fill and draw process. The follow-
ing figure (Figure 4-2) shows the effects of the initial
water age on the modeled results. As illustrated, a
good initial estimate for water age (120 hours in this
case) results in a much shorter time period until the
dynamic equilibrium is reached. In fact, in this case
when the initial age was input as zero hours, the model
did not even come close to reaching dynamic equilib-
rium during the simulation period and would have
required a much longer run duration to reach the same
point.
Model Application
Water age
Source tracing
Conservative constituent
Reactive constituent
Initial
Conditions
YES
YES
YES
YES
Reaction
Coefficients
NO
NO
NO
YES
Source
Quality
Usually NO
Usually NO
YES
YES
Initial age 0 hours
Initial age 120 hours
Figure 4-2, Effects of the Initial Water Age on the
Modeled Results.
• Source tracing: No explicit water quality
calibration can be performed because there are
no reaction coefficients. Estimates of initial
conditions in tanks for percentage of water
coming from a source are desirable in order to
shorten the length of the simulation. Values for
sources are usually set to zero for all sources
except for the specific source being traced.
• Conservative constituents: No explicit water
quality calibration can be performed because
there are no reaction coefficients. Estimates of
initial conditions in tanks for concentrations of
the conservative constituents can usually be
determined from field data and are desirable in
order to shorten the length of the simulation.
Values for sources are set to the typical
concentrations found in the source.
• Reactive constituents: For reactive constituents,
both the form of the reaction equation and the
reaction coefficients must be provided. When
modeling chlorine or chloramine decay, the
most common formulation is a first order decay
equation including both bulk and wall decay
coefficients. Values for these coefficients
typically require laboratory and field analysis
and calibration in order to match model results
to the concentrations measured in the field.
Correspondingly, THMs, a group of DBFs
formed when water is chlorinated or
chloraminated, generally increase in
concentration with time (Vasconcelos et al.,
1996). This process is frequently represented as
a first order growth function that asymptotically
approaches a limiting value representative of
maximum concentration reached when all of the
NOM has reacted or all of the chlorine has been
consumed. Both the limiting value and the rate
of growth must be determined in this case.
Water quality modeling is very sensitive to the
hydraulic representation of the system. To reiterate,
hydraulic calibration that may be sufficient for some
hydraulic simulation may require additional calibra-
tion when used as a basis for water quality modeling.
4-3
-------
A Reference Guide for Utilities
4.2 Static Calibration and
Dynamic Calibration
Just as water distribution system models can be run
in a steady-state or an extended period mode,
calibration can be performed in either a static mode
using a steady-state model or in a dynamic mode
using an extended period model. A common
approach is to perform a static calibration first
followed by EPS, to enhance the static calibration
through a dynamic calibration. The options and
procedures for these two types of calibration are
described below.
4.2.1 Steady-State Calibration Methods
The two most common approaches used in calibrat-
ing a steady-state hydraulic model are C-factor tests
and fire-flow tests. For water quality models of
chlorine/chloramines, a test procedure for estimating
bulk and wall demand may be employed. In all of
these cases, field data is collected under controlled
conditions and then applied to determine the model
parameters that result in the best fit of the model to
the field data.
4.2.1.1 C-Factor Tests
C-factor tests (sometimes called head loss tests) are
performed to estimate the appropriate C-factors to be
used in a hydraulic model. The C-factor represents
the roughness of the pipe in the widely used Hazen-
Williams friction equation. Typically, such tests are
performed on a set of pipes that are representative of
the range of pipe materials, pipe age, and pipe
diameters found in the water system that is being
studied. The results of the tests are then used to assign
C-factors for other pipes of similar characteristics.
In a field test, a homogeneous section of pipe between
400 and 1,200 feet long is initially isolated. Subse-
quently, flow, pipe length, and head loss are measured
in the field. Typically, nominal pipe diameters are
The underlying concept for a C-factor test is that all
factors in the Hazen-Williams friction equation can be
measured in the field and the equation can then be
solved for the unknown C-factor. It can also be used
to account for minor losses that occur through distri-
bution system components (e.g., valves, fittings). The
following equation is the Hazen-Williams equation
(Equation 2-3) arranged to solve for roughness.
C = 8.71 V D-°63 (H/L)-054 (Equation 4-1)
where
C = roughness factor
V = velocity in feet per second
D = pipe diameter in inches
H = head loss in feet
L = pipe length in feet
Flowed
Fire Hydrant#1
Fire Hydrant #2 HYdrant
Pitot Gage
Length
Closed Valve -""^
Figure 4-3. Schematic of Standard Two-Gage C-
Factor Test Setup.
Differential
pressure gage Flowed
Fire Hydrant #1 <9 Fire Hydrant #2 HVdrant
n-. . „
Pitot Gage
Length
^f
Closed Valve -"""^
Figure 4-4. Schematic of Parallel Hose C-
Factor Test Setup.
taken from system maps and these values are used
along with flow rate to calculate velocity. There are
two alternative methods for determining head loss in
the pipe section: a two-gage method (Figure 4-3) and
a parallel hose method (Figure 4-4). With the two-
gage method, pressure is read at hydrants located at
the upstream and downstream end of the section and
used along with elevation difference between the ends
to calculate head loss. With the parallel hose method,
a small-diameter hose is used to connect the two
hydrants to a differential pressure gage to directly
measure the difference in pressure. The two end
hydrants should be spaced far enough apart and there
should be sufficient flow so that there is a pressure
drop of at least 15 pounds per square inch (psi) for a
two-gage test or a 3-psi pressure drop for a parallel
hose test (McEnroe et al., 1989). In both cases, a
hydrant downstream of the test section is opened to
induce flow and a sufficient pressure drop. Multiple
downstream hydrants may also be employed to induce
a greater flow and larger pressure drop. Typically, a
pitot gage (as shown in Figures 4-3 and 4-4) is
attached to the flowing hydrants to measure the flow
rate. It is important to ensure that all flow between
hydrants is accounted for (i.e., any connections that
may bleed water into or out of the test section). The
two-gage method is the more commonly used
approach. The parallel hose method requires more
4-4
-------
A Reference Guide for Utilities
specialized equipment, but is inherently more
accurate and may be used when a large pressure drop
cannot be achieved. Note that the valve is closed
downstream of the flowing hydrant.
As noted above, an assumption is made that the pipe
diameter has not diminished from its original nominal
diameter due to tuberculation on the pipe walls. If
that assumption is not valid, the calculated C-factors
will be lower than expected. If very low C-factors are
calculated based on a field C-factor test, it is recom-
mended that further actions be taken in order to
determine the effective diameter of representative
pipes. These actions could include direct inspection
of sample pipes or use of calipers inserted into the
pipe to measure the effective pipe diameter.
4.2.1.2 Fire-Flow Tests
Fire-flow tests are routinely performed by water
utilities to determine the ability of the system to
deliver large flows needed to fight fires. In such a
test, fire hydrants are opened, the flow through the
hydrants measured and pressures measured at adjacent
hydrants (see Figure 4-5). The high demands caused
by the open hydrants lead to high flows and increased
head loss in pipes in the area around the hydrants.
Under these conditions, the system is stressed and the
capacity of the system to deliver these flows is very
sensitive to the roughness of the pipes.
These fire-flow tests can also be very effective as a
calibration methodology. In this case, in addition to
the standard information routinely collected as part of
a fire-flow test (flows and pressures), information is
collected on the general state of the system such as
pump and valve operation, tank water levels, and
general system demand. The distribution system
model is then run under the system conditions
observed during the test and adjustments made in
roughness factors (or other parameters) so that the model
adequately represents the data measured in the field.
P=42psi
P=36psi
Q=400gpm
Figure 4-5. Fire-Flow Test Setup.
P = 55 psi
P = 52 psi
Figure 4-6. A Hydrant Being Flowed with a Diffuser
as Part of a Fire-Flow Test.
Figure 4-6 illustrates an example setup for a fire-flow
test. The diffuser attached to the hydrant in the figure
includes a pitot gage used to measure the flow. The
cage diffuses the flow and prevents any objects in the
stream from being projected out at high speed.
In the case shown in Figure 4-5, only a single hydrant
is opened, with the flow measured at that hydrant and
pressure measurements made at four hydrants.
Additional hydrants may be flowed and monitored as
part of a fire-flow test for calibration (see Case Study
in Section 7.7).
4.2.1.3 Chlorine Decay Tests
Chlorine bulk reaction and wall reaction (or demand)
testing procedures can be used to determine the
reaction parameters used in water quality models.
Bottle tests measure the rate of chlorine reaction that
occurs in the bulk flow independent of wall effects.
This procedure is performed by first measuring the
chlorine at a representative location such as in the
effluent from a water treatment plant. Then several
bottles are filled with the same water and kept at a
constant temperature. Separate bottles are subse-
quently opened at intervals of several hours (or days)
and the chlorine content is measured. The resulting
record of chlorine at different times is used to estimate
the bulk reaction rate. See AWWA (2004) for a more
complete protocol for this test.
The purpose of the chlorine decay field testing
procedure is to estimate the chlorine wall demand
coefficient for representative pipes in the distribution
system. The method described here involves the
measurement of chlorine concentrations in a pipe
segment under controlled flow conditions and use of
the resulting chlorine measurements to determine the
wall reaction rate for that pipe segment. The method
is designed to be complementary with C-factor testing
so that it can be conducted in conjunction with a C-
factor test. The method is considered to be experimental
4-5
-------
A Reference Guide for Utilities
and feasible only for pipes that are expected to have
relatively high wall reaction values, such as smaller
diameter unlined cast iron pipes. For the smaller
diameter unlined cast iron pipes, pipe sections with a
length in the range of 1,500 to 2,000 feet will be
required to estimate wall demand. For other types of
pipes that typically have low wall decay factors (e.g.,
plastic and new pipes), the required length of the pipe
may be so long as to make this test impractical. Other
factors that should be considered in selecting sites
include the following:
• Ability to measure flow in the pipe.
• Ability to valve off the pipe segments.
• Presence of a reasonable chlorine residual
(preferably > 0.4 mg/L) at the upstream end of
the pipe segment.
• Ability to vary flow in the pipe over a
reasonable flow range (e.g., for a 6" pipe, a
range of flows of 100 to 500 gpm would be
desirable).
• Ability to estimate the actual pipe diameter for
the pipe segment.
For the selected pipe segment, major lateral(s) and
downstream segments should be valved off to control
flow in the pipe. Two or three sampling points should
be established along the segment of interest (up-
stream, downstream, and an optional midpoint).
Typically, these would be taps on fire hydrants. Prior
to the testing, the taps should be run for several
minutes to clean out the line. The approximate travel
time through the pipe should be calculated and
chlorine measurements taken from upstream to
downstream so that approximately the same parcel of
water is sampled at each station. Flow measurements
can be made at any location within the segment.
The test should be repeated for three flow values: a
low flow rate, a medium flow rate, and a high flow
rate. During each flow test, chlorine residual should
be measured at each of the two or three sampling
points. Since relatively small variations in chlorine
concentration are expected, a good quality field
chlorine meter should be employed and three repli-
cates should be taken at each sampling point for each
flow test. Following the field analysis, a spreadsheet
can be used to back calculate the resulting wall
reaction coefficients, or a water distribution model
can be used to determine the wall reaction coefficient
through trial and error.
4.2.2 Dynamic Calibration Methods
Dynamic calibration methods are associated with the
use of an EPS model. The dynamic calibration
methods include: (1) comparison of modeled results
If measured and modeled records of tank water levels
do not agree well, the relationship between the two
traces can provide clues as to the potential problems.
In the example depicted below, the timing of the fill
and draw cycles in the measured and modeled results
are quite close but the modeled and measured depth of
the fill cycles vary significantly. This suggests that the
system demands may be in error, resulting in an
incorrect amount of flow entering the tank.
Water
Level
Modeled
Time
In the second example illustrated below, the magnitude
of the change in water level is quite close in the
modeled and measured results, but the timing of the fill
and draw cycles differ. This is typically caused by
errors in the pumping controls in the model, resulting
in pumps being turned on and off at the wrong time.
Water
Level
Measured
Time
to measurements made in the field over time, and (2)
tracer studies. In both cases, model parameters are
adjusted so that the model adequately reproduces the
observed behavior in the field. Tracer studies are
discussed in detail in Chapter 3.
Comparison of modeled and measured data can be
used for calibration of both hydraulic and water
quality models. The most commonly measured
hydraulic data are tank water levels, flows, and
pressures. Frequently, this information is routinely
reported through SCADA systems to a database and
can be extracted. In other cases, continuous flow
meters or pressure gages must be installed to collect
data during a test period. Generally, tank water level
data and flow measurements are the most useful form
of data for calibrating an extended period model.
Under average water use conditions, temporal
variations in pressure measurements typically vary
over a relatively small range and then only in
response to variations in tank water levels. As a result,
they are less useful in calibrating model parameters.
If pressure measurements are going to be used for
4-6
-------
A Reference Guide for Utilities
dynamic calibration, the system must be stressed by
conducting fire-flow tests during the testing period.
The primary model parameters that are adjusted
during dynamic calibration are: demand patterns,
pump schedules and pump curves, control valve
settings, and the position (open or closed) of isolation
valves.
Dynamic calibration procedures using tracer study
data is discussed via a case study in Section 4.4 of
this chapter. Dynamic calibration can also be used for
calibrating water quality parameters, such as the wall
demand coefficient for computing chlorine residuals.
Generally, water quality field studies are performed in
conjunction with field hydraulic studies or with a
tracer study. For chlorine models, measurements of
chlorine are taken at frequent intervals in the field at
representative sites. These may include dedicated
sampling taps, hydrants, tank inlet/outlets, or other
accessible sites. Continuous chlorine meters may also
be used. During the model calibration process, the
model is first calibrated for hydraulic parameters, and
water quality coefficients are subsequently adjusted
so that the model results match the field data.
4.3 Manual Calibration and
Automated Calibration
The aforementioned process of adjusting model
parameters so that the model reproduces the hydraulic
and/or water quality results measured in the field can
involve a significant amount of effort in large or
complex systems. As discussed earlier in this chapter,
there are many parameters that can be adjusted in the
model and the combinations of possible parameter
values can sometimes appear to be quite overwhelm-
ing. Typically, a manual trial and error approach is
used. The most influential parameters can be identi-
fied based on sensitivity analysis and then adjusted to
see if they improve the results. This process is
continued until an acceptable level of calibration is
achieved or until budgetary constraints dictate
closure. It is not unusual for many (dozens or even
hundreds) separate model runs to be made in this
process.
An extension to the manual calibration process is an
automated approach that allows the computer to
search through different combinations of model
parameters (with a realm of realistic values) and to
select the set of parameters that results in the best
match between measured and modeled results. The
development of this type of program has been the
topic of many studies over the past 25 years (Walski
etal.,2003).
Automated methods require a formal definition of an
objective function for measuring how good a particu-
lar solution is. Generally, the value of a solution is
measured by a statistic that reflects the deviation
between measured and modeled results in flow and
pressure. A commonly used objective function is
minimization of the square root of the weighted
summation of the squares of the differences between
observed and predicted values. The weighting is used
to establish a relationship between the errors associ-
ated with flow and pressure. For example, the user
may choose a 1-psi error in pressure prediction to be
equivalent in value to a 10-gpm error in flow.
In most automated methods, the user also groups
pipes by common characteristics, such as age,
material, and nodes, into common demand character-
istics such as residential or commercial. The user then
specifies a range of allowable values for pipe rough-
ness factors or a range of multipliers applied to the
existing roughness factors. Similarly, a range of
allowable demand multipliers is also specified, as are
potential pipes where an existing isolation valve may
be closed. The optimization routine is then applied
and the roughness, demands, and isolation valve
positions are selected that result in the minimum error.
Though manual calibration still remains the predomi-
nant methodology, automated calibration methods are
becoming more available in commercial modeling
packages. It is likely that as the automated calibra-
tion methods are refined, the technology will expand
for routine use with EPS hydraulic and water quality
models.
4.4 Case Studies
In order to illustrate some of the calibration methods
described earlier in this chapter, two case studies are
presented in this subsection. The two case studies are
similar in general methodology but differ in the
overall scale and specifics of the study area. In both
cases, the distribution system model that was used as
a starting point for the calibration exercise was part of
a skeletonized model extracted from unspecified
portions of the GCWW distribution system.
Most larger urban water systems generally have at
least a skeletonized model of their distribution
system. It should be noted that (as discussed in
Chapters 2 and 3 of this report) a skeletonized model
denotes a model that includes only a major subset of
actual pipes rather than all pipes in the distribution
system. The extracted system model was modified
and converted to EPANET format for use in this
project. The modifications included: addition of key
pipes, updates to consumer demand data, and an
interconnection between the case study area and the
full system by a fixed grade node (reservoir). These
portions of the base model had been previously
calibrated using various dynamic calibration methods
and were used for routine water utility work. For the
4-7
-------
A Reference Guide for Utilities
purposes of calibration, separate field studies were
conducted in each study area.
In both field studies, a food-grade conservative tracer
(calcium chloride) was introduced into the system and
its movement through the system was monitored by
both grab sampling and continuous monitoring (CM)
stations installed at key locations in the distribution
systems. The CM stations were installed at hydrants
which were left partly open for the duration of the
study to minimize travel time between the main and
sampling location. Each open hydrant was added as a
new demand node in the EPANET network model.
Additionally, several ultrasonic flow meters were
installed to provide continuous flow measurements at
key locations. The general procedures, methodology,
and instrumentation used in these field studies are
consistent with those presented in Chapter 3.
4.4.1 Case 1 - Small-Suburban, Dead-End
System
This system is part of a larger pressure zone. It was
selected because of the relatively compact size and
simple structure, fed by a single supply pipe with no
additional storage. As a result, the movement of the
tracer was relatively rapid through the system and it
could be monitored with continuous meters placed at
several locations. The general layout of this sub-
system, the location of the injection site, and the mon-
itoring locations for this study are shown in Figure 4-7.
The calcium chloride tracer was injected as two
pulses, a two-hour pulse followed by a 2.5-hour
period of no injection and then followed by a higher
concentration pulse of two hours duration. The
injection rate and the resulting concentration of the
tracer in the distribution system just downstream of
Compliance with state and federal regulations during a
tracer study is obviously quite important. In order to
ensure that the tracer will not exceed allowable levels,
it is necessary to monitor information such as the rate
of injection of the tracer, the flow in the receiving pipe,
and the resulting concentration in the receiving pipe.
Frequently, a safety factor for the injection rate is
included to account for uncertainty. In this field study,
the tracer injection rate was very low and the flow
meter on the injection pump provided approximate
values. Chloride concentrations were monitored at a
suitable location approximately 100 feet downstream
of the injection point with a travel time of approxi-
mately 10 minutes. Due to unexpected variations in
flow through the pipe, delay in measurements, and
related computations (associated with tracer travel
time), chloride values exceeding the target level were
experienced for a brief period before the injection rate
was adjusted.
d>CM12
CM03 CM04
| Flow gage
O Conductivity meter (CM 1-20)
/\ Injection point
Figure 4-7. Schematic Representation of
Small-Suburban Dead-End System.
the injection point were carefully monitored to ensure
that the resulting chloride concentration did not
exceed the secondary maximum contaminant level
(MCL) of 250 mg/L for chloride.
The movements of the tracer pulses were monitored
by using both manual sampling and continuous
conductivity meters located throughout the distribu-
tion system. Additionally, four ultrasonic flow meters
were installed in the study area to provide continuous
flow measurements at key locations within the
distribution system.
In preparation for the calibration process, the conduc-
tivity readings were converted to chloride concentra-
tions using a relationship developed in the laboratory.
Figure 4-8 shows the relationship between conductiv-
ity and chloride and the best-fit linear and polyno-
mial relationships between them. This conversion was
necessary because conductivity is not a truly linear
parameter and, as a result, cannot be simulated
exactly in a water distribution system model. The
converted continuous concentration readings were
then compared to the manually collected data for
quality control purposes. Figure 4-9 shows the
resulting chloride data set that was used at one
location as a basis for evaluating model predictions as
part of the calibration process.
The preliminary results indicated some discrepancy
between the EPANET-model predicted values and the
400 600
Conductivity ijjs/cm)
Figure 4-8. Empirical Relationship Between
Chloride and Conductivity.
4-8
-------
A Reference Guide for Utilities
Elapsed time (hours)
Figure 4-9. Sample Chloride Data Used at One
Station for Calibration.
actual field-measured values, indicating the need for
model refinement and re-calibration to improve the
prediction capability of the EPANET model. There-
fore, EPANET modeling was performed to evaluate
the following four levels of model refinements:
• Level 1 (prior to calibration): A skeletonized
EPANET model was used with the original
hourly demand pattern provided by GCWW and
a time-step injection pattern of 60 minutes.
• Level 2: The same as Level 1, but a refined 10-
minute time-step pattern for injection was used
along with the conversion of the original hourly
demand patterns to 10-minute patterns.
• Level 3: The same as Level 2 with a refined
demand pattern for each node using the field-
measured flow data, addition of demand nodes
representing water demand of the partially open
hydrants, adjustment for a large industrial user
of water in the study area (based on data
obtained during the study), and the residential
water billing information provided by GCWW.
• Level 4: The same as Level 3 with a detailed
all-pipe (non-skeletonized) EPANET model.
The results of the four-stage model refinement and
calibration process are shown in Figure 4-10 for a
continuous monitoring location (CM-18) located on
the main feeder pipe. As illustrated, the improve-
ments in the demand estimates and inclusion of the
system details in the all-pipe model resulted in a vast
improvement in the model's prediction ability for that
monitoring location. Similar results were found for
most monitoring locations on the main pipe.
During the calibration and refinement process, various
model inputs such as flow, demand, and pipe
characteristics were adjusted to improve the model
prediction. The EPANET model was considered to be
calibrated for the area when the field data matched the
model-predicted output to an acceptable degree based
on visual observation. Depending upon the location
of the junction (where the model predictions are
compared with the field values), both concentration
and predicted time of tracer arrival might not be in
perfect agreement due to local variation in demands,
local flow velocities, and dilution impacts. The sharp
tracer fronts observed in this field study made it
difficult to employ quantitative statistical measures
(e.g., mean error, standard deviation, root mean square
error). Therefore, a graphical (visual) approach was
considered to be more suitable for model calibration
300.0
=3, 20°'°
500-
0.0
300.0
1
500-
0.0
300.0
"E
0
0.0
300.0
E.
1
0.0
Level 1
Computed • Observed ll —
,
™
D
/»»
n
V
0 4 8 12 16 20 24 28 32 36
T me (hours)
Level 2
,
_ .
• ^
n
r1
•/..
1
H
1
.1
T-
0 4 8 12 16 20 24 28 32 36
T me (hours)
Level 3
£
./•
<
A
U-
i
.
* * **
. i/1
u
0 4 8 12 16 20 24 28 32 36
T me (hours
Level 4
//
!A
P
i
L~J
u
KJL •
Wte
.
0 4 8 12 16 20 24 28 32 36
T me (hours)
Figure 4-10. Comparison of Model Versus Field Results
for Continuous Monitor Location CM-18 at Various
Calibration Stages.
4-9
-------
A Reference Guide for Utilities
in this application. For example, if the prediction of
the arrival time for the tracer differs by even a few
minutes from the observed arrival time, use of these
standard measures of error could result in a high
number, even though the prediction could be viewed
graphically as very good.
The calibration of the "looped" portion (referring to
the portion of the network on the bottom right hand
side of Figure 4-7) of this network proved to be more
difficult and the results for some monitoring locations
on the looped piping were less satisfactory. The most
problematic were continuous monitoring locations
CM-02 and CM-04. Monitoring station CM-02 was
located near the confluence of two separate loops,
with the actual monitored connection being slightly
offset from the junction node. Examination of the
model results showed that flow reached that junction
from both directions and small variations in the
amount of flow in each of the loops resulted in very
different travel times. As illustrated in Figure 4-11,
this complex travel pattern along with the offset
location of the monitoring station resulted in poor
300.0
250.0
5 200.0
o>
.§
g mo
1
g 100.0
50.0
0.0
300.0
250.0
5 200.0
M.
g 150.0
6 100.0
50.0
0.0
300.0
250.0
§ 200.0
.E.
g 150.0
g 100.0
50.0
0.0
CM-02
—\-
— Compute
d • Obs
•
• •
D
• ,
A
/ IK
* — 1
i
0 4 8 12 16 20 24 28
Tme (hours
CM-03
f
I
•
(i
•
.
n
32 36
0 4 8 12 16 20 24 28
Tme (hours
CM-04
/!
A.
r • •
M
I " 1
r 1
• •
32 36
0 4 8 12 16 20 24 28
Tme (hours
32 36
Figure 4-77. Calibration of "Looped Portion.
prediction of travel time to that station. Also,
monitoring station CM-04 is located at the end of a
dead-end pipe section and travel to this node is
strongly influenced by demands at the very far end of
the dead-end section. As illustrated in Figure 4-11,
this resulted in a poor match of the peak concentra-
tion during the second pulse. It is also postulated that
dispersion, which is not represented in EPANET, may
have had an influence on the peak concentration due
to the very low velocities in the dead end pipe. In
some cases, this could also be caused by inaccurate C-
factors as applied to the distribution system. However
(as illustrated in Figure 4-11), for monitoring location
CM-03 located in the main part of the looping system,
the model and field agreement was quite good.
Case 1 data illustrates that, depending upon the level
of refinement and calibration, there is a significant
variation in the capability of a model to accurately
represent the system. In general, the parts of the
network that are configured as trees (main stem with
branches) are more easily calibrated by making
adjustments in demands. For looping parts of the
system and at dead-ends, results are very sensitive to
small variations in demands and system configura-
tion, leading to the possibility of significant predic-
tion errors at some locations. Uncertainty in demand
estimates can be a major source of error in the model
estimates.
4.4.2 Case 2 - Large-Suburban Pressure Zone
Similar to Case 1, a field study and calibration
exercise was carried out in a large-suburban, pressure
zone. This area was selected in order to demonstrate
the application of tracer studies and calibration
techniques in a more complex area. The selected area
contained multiple pumps and tanks. The selected
distribution system area is representative of relatively
complex, well-gridded systems found in many larger
water systems. The layout of the system, the location
of the injection site, and the monitoring locations are
shown in Figure 4-12.
Two separate tracer studies were performed in this
zone. The first study was used to further calibrate the
skeletonized model received from the water utility.
The second study served as a validation event to test
the veracity of the calibrated model. In the calibration
event, the tracer was introduced directly into the main
feed line servicing the entire area (characterized by
higher flow/higher pressure). In the validation study,
the tracer was pulsed. A total of 34 continuous
conductivity meters were installed in the system.
Four flow meters were temporarily installed to provide
flow measurements at key locations.
During the calibration study, the calcium chloride
tracer was injected into the main feed line serving the
4-10
-------
A Reference Guide for Utilities
<> Injection Point
D Flow Monitoring Location
o Continuous Monitoring Sample Location
Figure 4-12. Schematic Representation of Case 2
Study Location.
area for a period of 6 hours. In the validation study,
the tracer was pulsed by fill and draw cycles in a
storage tank at the same location. In both cases, a
target chloride concentration of 190 mg/L or lower
was set in order to safely not exceed the 250 mg/L
secondary MCL for chloride.
During the calibration process, initial EPANET
model simulations were reviewed in detail to
determine the flow patterns around various moni-
toring locations and to attempt to identify causes
for discrepancies in the observed and predicted
values. A careful examination of the areas of
significant discrepancies indicated that these were
primarily limited to three geographic sub-regions
within the skeletonized network. In addition to
these three sub-regions, there were a few isolated
locations where the predicted tracer pattern did not
match the observed tracer pattern from the field
study. The modeling team carefully examined each
of these regions and addressed the zonal issues
accordingly. The three sub-regions are shown in
Figure 4-12.
In Region 1 (CM42, CM43, and CM44), the field data
indicated that the tracer arrived at these continuous
monitoring locations several hours before the model's
prediction. On closer inspection, it was found that a
potential flow path existed which was not included in
the skeletonized model. While the pipe diameter was
small, it significantly altered the hydraulic water flow
path to that region. This missing pipe-link was added
to the model, using the appropriate pipe parameters.
Furthermore, the modeling team investigated the GIS
database to see if there were any substantial changes
in these areas since the time when the original water
demand patterns were developed five years ago. The
updated GIS information indicated a presence of
recent housing development in that region. There-
fore, additional demand nodes were entered into the
EPANET model to accommodate for this develop-
ment. Another possibility for the discrepancy was
that the demand in this region was significantly
higher than the average residential demand modeled
in the area. To simulate this possibility, a sensitivity
analysis was performed in which the modeled demand
in this region was doubled. The model-predicted
results improved significantly for this region based on
these three adjustments.
In Region 2 (CM52, CM53, CM55 and CM56), an
opposite phenomenon to that in Region 1 was
observed. The field data indicated that the tracer
arrived several hours after the model's prediction.
One possible explanation was that this region had
lower demand than the average residential demand
modeled in this area. The flow meter data upstream of
this location supported this theory as the EPANET
predicted flow in this pipe was much higher than the
field observed flow (see Figure 4-13a). To simulate
this possibility, the local demand in this region was
reduced by 30 percent in the model. The resultant
flow matched the flow meter data (see Figure 4-13b).
Also, similar to Region 1, it was found that a potential
flow path had again been left out due to
skeletonization of the model, which affected CM52.
This pipe link was added to the model using the
appropriate pipe parameters. Distribution mains
between CM55 and CM53 were also found to have
been upgraded since the EPANET network model was
60
4-11
-------
A Reference Guide for Utilities
developed for this area. The EPANET model pipes for
this location were updated using the newer informa-
tion. The model-predicted results improved signifi-
cantly for this region based on these three adjustments.
In Region 3 (CM34 and CM35), the field data
indicated that the tracer arrived at locations CM34
and CM35 several hours after the model's predicted
arrival time. However, the field-verified tracer arrival
time matched the predicted tracer arrival time at
location CM33 which is slightly upstream of these
locations. Also, a review of the water flow pattern in
this region indicated that the water traveled from
CM33 towards CM34 and CM35 (at all times). Based
on the demands in the EPANET model, the pipe
lengths, and the regional water flow information, the
delay in tracer arrival at CM34 and CM35 could not
be explained. A closer inspection of the region
revealed a complex grid of interconnected pipes in
this region, which were skeletonized as two parallel
pipes. This skeletonization eliminated a number of
different possible hydraulic flow paths between CM
33 and CM34/CM35. Also, in the EPANET model
inputs, it appeared that the demand close to CM34
and CM35 was set artificially higher (to account for
the overall demand in the skeletonization process).
This model setup resulted in the predicted faster tracer
arrival times at CM34/CM35 than those observed in
the field. To account for this anomaly, a few pipe
segments from the master plan were added to the
skeletonized model of this region to better simulate
the actual grid demands near CM34 and CM35. This
model adjustment resulted in better prediction of the
tracer arrival times.
During the calibration process, as demands were
adjusted, a mass balance was performed for each hour
to ensure that the net water demand in the study area
remained the same, i.e., the increase in the demand at
certain nodes was balanced by the reduced demand at
other nodes to eliminate any net impact on water
demand. In the final refinements, a multiplier of 2.0
was used for the base demand in Region 1, and a
multiplier of 0.7 for the base demand in Region 2.
These refinements showed some improvement in the
model's ability to correctly predict the tracer arrival
time and concentration. These calibration efforts
resulted in a relatively well-calibrated network model.
However, some local problems remained, especially in
looped areas and areas that were branched off from the
main lines.
The substantial changes made to the EPANET
skeletonized model representing the large area
necessitated a validation process. Therefore, the
calibrated EPANET model input file from the first
event was used to validate the model's capability to
predict the results during the subsequent tracer
addition. For the purposes of this validation, the data
300.0
250.0
,-j 200.0
•a 150.0
g 100.0
50.0
0,0
E
/
f
E
— Co
3
• T
•
\ •
Tiputed
•
• Ob
• _ ,
served
_
II
0 5 10 15 20 25 30 35 40 45 50 55 60
Time (hours)
Figure 4-14a. Chloride Concentration for Calibration
Event at Continuous Monitor Location CM-59.
Computed
0 8 16 24 32 40 48 56 64 72 80 88
Time (hours)
Figure 4-14b. Chloride Concentration for Validation
Event at Continuous Monitor Location CM-59.
from the second set of pulsed injections was modeled
using the calibrated EPANET network model for the
study area to see how the predicted results compared
with the continuous monitoring data collected during
this event. The modeled and measured concentrations
are compared in Figure 4-14a for the EPANET
calibration. A similar comparison is shown in Figure
4-14b for the validation study.
Additionally for the purposes of this analysis, the
EPANET predictions from the validation event were
compared with the field results for each monitoring
site and each site was given a grade as follows:
• Very good match (within +20 percent of the
actual concentration and within +1 hour of the
actual tracer arrival time)
• Moderate match (within +30 percent of the
actual concentration and within ±5 hours of the
actual tracer arrival time)
• Poor match (greater than ±30 percent of the
actual concentration or greater than ±5 hours of
the actual tracer arrival time).
Of the 34 monitoring sites in this study area for the
validation event, 15 received a grade of very good
match, 14 were in the moderate match category, and 5
received the lowest grade of poor match. In general, it
was found that better matches occurred on larger pipes
serving large populations, while the poorest matches
occurred in more localized loops serving fewer
4-12
-------
A Reference Guide for Utilities
customers. These results are, in general, quite similar
to the results obtained for the calibration event, and
most problems repeatedly occurred at the same
locations for both events. The validation event
results confirm the fact that the calibrated EPANET
network model can now be used to predict the
outcome of a separate event to the same degree of
accuracy.
4.5 Future of Model Calibration
Calibration continues to be a major focus of most
modeling efforts. It can provide a model that may be
used with greater confidence and produce results that
are commensurate with the important decisions that
are made based on the application of the model.
However, there is significant room for improvements
in calibration methodologies and in developing a
standardized set of calibration protocols. This has led
to an active research program in this area that is
expected to continue into the future.
4.5.1 Calibration Standards
The following issues are raised frequently in the field
of distribution system modeling:
• extent of calibration needed for various
applications, and
• standards for calibration.
Though these are very reasonable questions, straight
forward answers are usually not readily available.
There is general agreement in the modeling profession
that the amount and degree of calibration required for
a model should depend upon the intended use of the
model (Engineering Computer Applications Commit-
tee [ECAC], 1999). Some applications such as design
and water quality analysis typically require a high
degree of calibration, while other uses, such as master
planning, can be performed with a model that has not
been calibrated to such a high standard. However,
there are no universally accepted standards.
In the United Kingdom, there are performance criteria
for modeling distribution systems (Water Authorities
Association and WRc, 1989). These are expressed in
terms of the ability to reproduce field-measured flows
and pressures within the model, as shown below.
Flow
1. +5 percent of measured flow when flows are
more than +10 percent of total demand
(transmission lines).
2. +10 percent of measured flow when flows are
less than ±10 percent of total demand
(distribution lines).
Pressure
1. 0.5 m (1.6 ft) or 5 percent of head loss for 85
percent of test measurements.
2. 0.75 m (2.31 ft) or 7.5 percent of head loss for
95 percent of test measurements.
3. 2 m (6.2 ft) or 15 percent of head loss for 100
percent of test measurements.
In 1999, the AWWA Engineering Computer Applica-
tions Committee developed and published a set of
draft criteria for modeling. These were not intended
as true calibration standards, but rather as a starting
point for discussion on modeling needs. These criteria
are summarized in the following table (Table 4-2).
Table 4-2. Draft Calibration Criteria for Modeling (based on ECAC, 1999)
Intended
Use
Long-Range
Planning
Design
Operations
Water
Quality
Level of
Detail
Low
Moderate to
High
Low to High
High
Type of
Simulation
Steady-State
or EPS
Steady-State
or EPS
Steady-State
or EPS
EPS
Number of
Pressure
Readings1
10% of Nodes
5% - 2% of
Nodes
10% -2% of
Nodes
2% of Nodes
Accuracy of
Pressure
Readings
+5 psi for
100% Readings
+2 psi for 90%
Readings
+2 psi for 90%
Readings
+3 psi for 70%
Readings
Number of
Flow
Readings
1% of Pipes
3% of Pipes
2% of Pipes
5% of Pipes
Accuracy of
Flow
Readings
+ 10%
+ 5%
+ 5%
+ 2%
1 The number of pressure readings is related to the level of detail as illustrated in the table below.
Level of Detail
Low
Moderate
High
Number of Pressure Readings
10% of Nodes
5% of Nodes
2% of Nodes
4-13
-------
A Reference Guide for Utilities
At this point, there is no clear movement toward
establishing calibration standards. However, it is
likely that the need for further guidance in this area
will increase as the extent and sophistication of
modeling continues to expand.
4.5.2 Technological Advances
Research is continuing in two areas that strongly
influence the likelihood of improved calibration of
water distribution systems models: monitoring
technology and optimization techniques. The
available optimization techniques (and those under
development) have been briefly discussed in this
chapter and in Chapter 2. Active research and
development areas include optimization techniques
for water quality calibration, EPS models, and use of
tracer data. Areas of research, development, and
experimental applications in monitoring technology
include less expensive meters that can be inserted
into pipes in the distribution system and automated
monitoring for use in conjunction with tracer studies
(as discussed in Chapter 3).
4.6 Summary and Conclusions
Water distribution system models can be used for a
number of purposes. Many of these uses result in
engineering decisions that involve significant
investments. It is therefore important that the model
represent the "real world." Calibration techniques can
be used to ensure that the mathematical representation
of the system, or model, adequately simulates the
system.
Calibrating a model is a difficult task because there
are many parameters that can be adjusted and finding
the combination of parameters that result in the best
agreement between measured and modeled results is
often challenging. It is recommended that the model
be calibrated using one set or more of field data and
subsequently validated with an independent set of
field data.
Calibration of water distribution system models can
be viewed in many dimensions. Hydraulic calibration
is used to adjust the parameters associated with
hydraulic simulations, while water quality calibration
is applied to reaction rates and other parameters that
control the water quality simulation. Static or steady-
state calibration methods are used with steady-state
models and data collected at instantaneous snapshots
in time, while dynamic calibration is conducted with
extended-period simulation models and time-series
data. Manual calibration techniques involve manual
application of models in a trial-and-error mode, while
automated calibration uses the power of the computer
to search a wide range of solutions and to select the
set of parameters that best achieve a stated objective.
Automated methods can reduce much of the tedium
During the calibration process, it is important to
eliminate various sources of errors in modeling. As a
first pass, a modeler should check for typographical
errors, accuracy of affected piping layout and material,
general system flow, velocity values, and distribution
system demands. Thereafter, one should look into other
sources of errors such as skeletonization, valve posi-
tion, geometric node placement anomalies, SCADA
data errors, and pump performance.
associated with calibration but require the modeler to
formally define a quantitative objective function for
measuring how well the model matches the field data.
Such automated methods are becoming more avail-
able in commercial modeling packages.
Two case studies are presented in this chapter. The
case studies differ in terms of the overall scale of the
study area. In both cases, the distribution system
model that was used as a starting point for the
calibration exercise was part of a skeletonized model.
The results demonstrate the need for adequate model
calibration.
The extent of calibration and calibration techniques
are a major issue in most modeling efforts. There is
significant potential for improvements in calibration
methodologies and in standardization of calibration.
This has led to an active and continuing research
program in this important area.
4-14
-------
A Reference Guide for Utilities
References
AWWA. Computer Modeling of Water Distribution
Systems (M32). AWWA, Denver, CO. 2004.
ECAC. "Calibration Guidelines for Water Distribu-
tion System Modeling." Proceedings, AWWA ImTech
Conference, New Orleans, LA. April 18-21, 1999.
McEnroe, B.M., D.V. Chase, and W.W Sharp. "Field
Testing Water Mains to Determine Carrying Capac-
ity." Miscellaneous Paper EL-89, U.S. Army Engineer
Waterways Experiment Station. Vicksburg, MS.
1989.
Vasconcelos, J.J., L.A. Rossman, WM. Grayman, P.P.
Boulos, and R.M. Clark. Characterization and
Modeling of Chlorine Decay in Distribution Systems.
AWWA and AwwaRF, Denver, CO. 1996.
Walski, T. M, D.V. Chase, D.A. Savic, W.M. Grayman,
S. Beckwith, and E. Koelle. Advanced Water Distri-
bution Modeling and Management. Haestad Press,
Waterbury, CT. pp 268-278. 2003.
Water Authorities Association and WRc. Network
Analysis - A Code of Practice. WRc, Swindon, UK.
1989.
Williams, G.S., and A. Hazen. Hydraulic Tables. John
Wiley & Sons, NY. 1920.
4-15
-------
A Reference Guide for Utilities
Chapter 5
Monitoring Distribution System Water Quality
Monitoring a water supply system and its various
components facilitates the gathering of data about the
state of the system (physical, operational and water
quality). If the state of the system has minimal
changes in time or space, a simple monitoring system
may be sufficient to define and manage the system
characteristics. However, if there is potential for
significant variation in the state of the system, the
monitoring system must be adequately designed to
capture that variability. Thereafter, depending upon
the type and magnitude of variability, an appropriate
response can be provided to restore the "normal"
system state. This chapter will focus on monitoring
water quality-related parameters in a distribution
system.
In a distribution system, water quality may vary due
to factors such as normal patterns in water consump-
tion, seasonal variations, source water quality,
components of the distribution system, operation of
the system, retention time in storage, travel time in
the piping system, or the condition of the system
itself. Variability may also result from unusual
occurrences, such as intentional/accidental intrusions
of contaminants, or chemical processes such as
nitrification. Design of a water quality monitoring
program must take into account both the nature of the
variability and the manner in which monitoring data
will be used. In other words, the objective of the
monitoring program must be defined along with
appropriate output or reporting requirements.
In general, monitoring systems can be defined based
on the uses or needs of the monitoring program, the
general type of monitoring to be performed (manual
grab sampling and/or continuous automated online
monitoring), or the specific monitoring equipment
characteristics. It is important to first establish a clear
objective(s) for monitoring. Thereafter, depending
upon the availability of funding, need, and expertise,
one should select the appropriate sampling
technique(s) and monitoring equipment. Once an
appropriate monitoring system has been selected and
implemented, it is important to operate and maintain
the program to achieve optimal results and benefits.
However, the system should be flexible enough so
that it can be modified in case it does not meet the
original objective(s).
This chapter discusses the various drivers or objec-
tives for monitoring followed by a summary of
available monitoring techniques. An overview of
monitoring equipment is presented followed by
guidelines for establishing monitoring requirements
(e.g., selection of parameters, number and locations of
monitors, and monitor characteristics). Some guid-
ance for engineering and evaluating remote monitor-
ing systems is also presented, along with some EPA-
sponsored monitoring case studies. The chapter
concludes with a summary and a listing of references.
The recent studies involving the use of online continu-
ous monitoring systems have resulted in large streams
of data that document the minute-by-minute changes
in water quality that exist at various points in the
water networks. The application of this technology
has the potential for providing new insights as to how
water distribution systems may be operated and
designed to improve water quality. However, these
systems will require a relatively high level of sophisti-
cation in terms of data management, including the
capability to generate real-time reports, graphical and
visual representation of information, and compliance
reports for meeting drinking water standards. Some of
these data streams may well reveal excursions in water
quality that constitute violations of current or future
drinking water standards, or a security-related inci-
dent. This type of information may put pressure on
drinking water utilities and regulatory agencies to take
remedial action, possibly on an emergency basis, even
when such actions may not be fully justified (or
warranted). However, careful planning and negotia-
tions with appropriate regulatory authorities to define
these potential "excursions" and the proper corrective
action to be taken would prevent any misunderstand-
ings and minimize or eliminate the potential for
unjustified enforcement or response actions.
5.1 Establishing Monitoring
Objective(s)
In order to define and implement an effective monitor-
ing plan, clear objectives must be established.
Collecting data just for the sake of accumulating
information is not cost effective. In drinking water
systems, there are several specific reasons to collect
data and, typically, the monitoring system is tailored
to meet one or more of these needs. The objectives of
monitoring distribution systems can be broadly
classified into the following five uses:
• regulatory driven monitoring,
• security related monitoring,
• process control related monitoring,
5-1
-------
A Reference Guide for Utilities
Table 5-1. Federal Distribution System Water Quality Monitoring Requirements
Regulation
TCR
SWTR and IESWTR
LCR
DBPR2
Monitoring Requirement(s)
Samples must be collected at sites that are representative of the water quality
throughout the distribution system based on a site plan that is subject to review by
the primacy regulatory agency.
The minimum number of samples that must be collected per month depends on the
population served by the system.
For each positive total coliform sample, there are repeat sampling requirements,
additional analyses, and an increased number of routine samples.
Disinfectant residuals must be measured at TCR monitoring sites.
All systems serving a population > 50,000 people must do water quality parameter
(WQP) monitoring.
The number of sample sites for Pb/Cu and WQP monitoring is based on system size.
The IDSE requirement of DBPR2 in turn requires the establishment of a Standard
Monitoring Program (SMP). The SMP will require one year of data on THMs and
Haloacetic Acids (HAAs). The number of sampling locations is based on utility size
and source characteristics. Modeling can reduce sampling requirements.
• water quality characterization (e.g., general,
baseline, or other research-related monitoring),
and
• multi-purpose (a combination of above) use of
monitoring data.
The following subsections present the overall scope
of each of these five objectives.
5.1.1 Regulatory Driven Monitoring
Various federal, state, or other governmental agencies
have regulations that specify distribution system
monitoring requirements. An overall review of federal
regulations impacting distribution systems was
presented in Chapter 1. The specific federal distribu-
tion system monitoring requirements (existing and
proposed) are summarized in Table 5-1. In some
cases, states have imposed more stringent criteria and
monitoring requirements.
5.1.2 Security Related Monitoring
Assessments performed by utilities and various
research studies have identified that water distribu-
tion systems are vulnerable to intentional (or acciden-
tal) contamination. In addition to "hardening"
systems in order to deter intentional contamination,
monitoring as part of an early warning system (EWS)
has emerged as a logical approach to cope with
potential contamination events. There are no existing
or proposed standards for such monitoring. However,
it is well recognized that monitors will need to be
sufficiently sensitive to a broad range of potential
contaminants and appropriately located to detect a
contamination event within a reasonable time.
Additionally, as detailed in EPA's Response Protocol
Toolbox (EPA, 2003-2004), monitors must be an
integral part of an emergency response management
plan in order to be effective. Extensive research and
development is underway on monitor development,
calibration, and placement in response to the per-
ceived security monitoring needs.
Currently, EPA has an ongoing test program to evalu-
ate the potential of sensors monitoring routine online
water quality parameters, such as pH, oxidation
reduction potential (ORP), free chlorine, total organic
carbon (TOC), conductivity, and turbidity, to serve as
rapid detection devices for detecting contamination
events in distribution systems. Online monitors were
selected because response time is critical for achieving
the objective of providing early warning. Both bench-
and pilot-scale studies are being conducted at the
Water Awareness Technology Evaluation Research and
Security (WATERS) Center within the EPA's Test and
Evaluation (T&E) Facility in Cincinnati, Ohio. The
bench-scale runs are designed to identify the detection
threshold of each sensor for specific contaminants.
The pilot-scale runs are designed to evaluate overall
response of the selected sensors by injecting known
quantities of potential contaminants into the distribu-
tion system simulator (DSS). For this purpose, several
pilot-scale DSSs have been fabricated and used for
these test runs. The sensor data are collected continu-
ously and archived electronically to establish stable
baseline conditions and to also record sensor responses
to injected contaminants. Grab samples are collected
periodically before and after injection of contaminants
to confirm the sensor results.
5-2
-------
A Reference Guide for Utilities
5.1.3 Process Control-Related Monitoring
Monitors can also be used in a distribution system to
provide real-time or near real-time information on
water quality that can then be used to control treat-
ment processes at a treatment plant or in the distribu-
tion system. The use of continuous chlorine monitors
in the distribution system to control disinfectant feed
rates at the plant or at in-distribution system booster
chlorination stations are examples of this type of
monitoring (Uber et al., 2003).
5.1.4 Water Quality Characterization
Information from long-term monitoring of distribution
systems can be used to develop baseline trends in
water quality for that system. Such information is
useful in evaluating a water supply system and for
planning upgrades or modifications to system design
or operation.
Additionally, if this information is appropriately
distributed, it builds consumer confidence and helps
to keep customers up to date about the water quality
so that they can use this information to make deci-
sions about protecting their health. Currently, there
are no standards or guidelines for this type of monitor-
ing. However, for this information to be useful and
cost-effective, a regular program for examining and
analyzing the collected information is essential.
5.1.5 Multi-Purpose Use of Monitoring Data
Monitoring can be an expensive undertaking in terms
of capital costs, as well as operation and maintenance
(O&M) costs, including labor. Costs include the
purchase and upkeep of equipment, laboratory
analysis, labor, and consumable supplies. The
investment in monitoring and automated monitoring
systems is justifiable if the resulting data are used for
more than one objective. For example, if data
collected for security purposes can also be used for
process control, it should be easy to justify poten-
tially large investments in automated monitoring
equipment. Monitoring systems should be properly
designed in order to meet multi-purpose requirements.
5.2 Monitoring Techniques
The two major factors in designing and implementing
an effective monitoring program are sampling
techniques and equipment selection. This section
focuses on available monitoring techniques. Samples
can be collected and analyzed in two ways: grab
samples and/or by automated online monitoring.
Automated monitors (continuous or discrete) are
sometimes supplemented with automated samplers
that can collect both discrete and composite water
samples for further analysis at a later date/time. Grab
samples are collected manually and analyzed in the
field or in the laboratory. Grab samples are labor-
intensive in comparison to automated sampling and
provide snapshot information about the system at the
time of sample collection. Automated monitoring
uses online instrumentation, and data is collected by
means of sensors and automated data loggers. They
can also be tied to a SCADA System. High-end
monitors require a higher capital expense for the
purchase and maintenance of sensors, data acquisi-
tion, data communication, data storage, and data-
processing hardware and software. However, this type
of monitoring provides a continuous time-series
profile of changes in water quality. Both automated
and grab sampling can be incorporated into a compre-
hensive monitoring plan. These techniques are
further discussed in the following subsections.
5.2.1 Manual Grab Sampling
Historically, routine water quality monitoring in
distribution systems has been carried out through
manual grab samples followed by analysis in the field
or in the laboratory. Essentially, all regulatory
monitoring is still carried out by this method. For
example, samples required for large community water
supply systems under the SWTR are manually
The equipment routinely required in a manual grab
sampling program includes field sampling equipment
(e.g., chlorine meter), safety equipment (vests, rain gear,
and flashlights), and laboratory equipment. Consum-
able supplies include sampling containers, reagents,
and marking pens. One should identify the needs and
availability of equipment and supplies and investigate
various sources for equipment. Because equipment
malfunction or loss is possible, some redundancy in
equipment is appropriate. Some important functions to
consider when establishing a field sampling program
include the following:
• Establish a systematic and organized method for
all sampling and data recording. Take notes to
document all aspects of the process.
• Provide training to sampling crews and specify
these training requirements in the sampling
program plan.
• Contingency planning is important; therefore,
consider the potential for equipment
malfunction, illness of crew members,
communication problems, severe weather,
malfunction, and customer complaints.
• Establish a communications protocol to
coordinate actions. A means of communication
is needed to respond to unexpected events.
Alternatives include radios, cellular phones,
walkie-talkies, or a coordinator in a vehicle to
circulate among field crews.
• Calibrate field analytical equipment before and
during the sampling activity.
5-3
-------
A Reference Guide for Utilities
collected at sites within the distribution system and
tested for disinfectant levels in the field. Samples taken
to satisfy the requirements of the TCR are also manually
collected in the field and subsequently analyzed in the
laboratory. Manual sampling is labor-intensive and the
number of samples that can be collected is limited by
availability of personnel and analysis costs. However,
they are specified by some regulations. Potentially
important events that may occur between the routine
grab samples may be lost (e.g., process upset). Also,
there is a potential for dismissing unusual grab
sampling results as some type of manual monitoring
error (Hargesheimer et al., 2002).
5.2.2 Automated/Online Monitoring
As stated in the report, "Online Monitoring for
Drinking Water Utilities" (Hargesheimer et al., 2002),
"There is an evolution from grab-sample monitoring
to online monitoring as sampling, analysis, data
processing, and control functions become more
automated." Online monitoring requires a mechanism
for moving the sample water from the distribution
system to an instrument, appropriate instrumentation
for analyzing the water, a mechanism for communicat-
ing the results, and a means of assessing the results of
the monitoring. Additionally, the instrumentation
must be periodically calibrated and maintained for
quality control/quality assurance.
In the past, distribution system online monitors were
typically housed in a controlled environment with
sample lines from the distribution system to the
instrument. This resulted in most instrumentation
being located at facilities such as tanks and pump
stations. The instrumentation was sometimes con-
nected to a SCADA system so that results could be
communicated to a central office. More recently,
some instrumentation is available that is designed for
installation in manholes or for direct insertion into
water mains.
The American Society of Civil Engineers (ASCE), in
concert with other leading organizations, entered into
a cooperative agreement with the EPA to develop
standards documents and guidance aimed at enhancing
the physical security of the nation's water and waste-
water/stormwater systems. Under this agreement,
ASCE is leading the effort to develop guidelines for
designing an online contaminant monitoring system
(OCMS). The Interim Voluntary Guidelines for
Designing an OCMS were published in December
2004 (ASCE, 2004). This document provides compre-
hensive information on several topics, including
rationale for OCMS and system design basics, selec-
tion and siting of instruments, data analysis, and use of
distribution system models.
5.3 Monitoring Equipment
Overview
In general, monitors can be categorized by the types
of parameters (contaminants, agents, and characteris-
tics) that the monitor is used to measure. For estab-
lishing water quality, the monitors are designed to
measure one or more parameters that represent
physical, chemical, and/or biological characteristics
of the system. Typically, in manual grab sampling
programs, hand-held physical and/or chemical
parameter measuring devices are used. These hand-
held devices are carried to the sampling location
along with appropriate containers to collect water
samples for performing more complex chemical and
biological analyses in a laboratory. The online
sampling devices are more complex devices that are
designed to automatically measure, record, and
display specific physical, chemical, or biological
parameters. A brief overview of these devices is
presented in the following subsections.
5.3.1 Physical Monitors
Physical monitors are used to measure the physical
characteristics of the water in a distribution system.
They include a variety of instrumentation that
measures various macro characteristics, such as flow,
velocity, water level, pressure, and other intrinsic
physical characteristics. Examples of intrinsic
physical characteristics include pH, turbidity, color,
conductivity, hardness, alkalinity, radioactivity,
temperature, fluorescence, UV254, and ORP. In
general, physical monitors tend to be relatively
inexpensive, quite durable, and readily available.
5.3.2 Chemical Monitors
Chemical monitors are used to detect and measure
inorganic or organic chemicals that may be present in
the water. A wide range of chemicals may be of
interest, and a large variety of technologies can be
used. A specific technology or multiple technologies
must be properly selected for a particular chemical or
a group of chemicals of interest. Examples of
chemical monitors include, but are not limited to
residual chlorine monitor, TOC analyzer, and gas
chromatograph/mass spectrometer (GC/MS). Typi-
cally, the same general type of technology may be
available in either automated online monitoring
capability or to support manual grab sample analysis.
5.3.3 Biological Monitors
Biological monitors (bio monitors) include bio-
sensors and bio-sentinels. Bio-sensors detect the
presence of biological species of concern, such as
some forms of algae or pathogens. The general
operating principles of bio-sensors may include
photometry, enzymatic, and/or some form of bio-
chemical reaction. The bio-sentinels use biological
5-4
-------
A Reference Guide for Utilities
organisms as sentinels to determine the likely
presence of chemical toxicity in a water sample.
In general, bio-sentinels cannot be used to identify
the presence of a specific toxic contaminant - rather
only that there is some form of toxic contaminant
present. Most bio-sentinels operate by observing the
behavior of selected organisms. Examples of such
organisms include: fish, mussels, daphnia, het-
erotrophic bacteria, and algae. When the sentinel
organism senses the presence of toxicity, it reacts in
some unusual manner. Bio-sentinel instruments
respond to these reactions and note that an unusual
event is occurring. This application is somewhat
analogous to the use of indicator organisms (e.g., total
coliforms) to indicate the water quality in the distri-
bution system.
While bio-sensors can be directly applied in distribu-
tion systems without pretreatment of the sample, the
bio-sentinels are typically used in source waters. This
is because most organisms are sensitive to the
presence of chlorine (or other disinfectants) in the
water. Therefore, if a bio-sentinel is proposed to be
used for distribution system monitoring, the water
must be de-chlorinated prior to entering the bio-
sentinel instrument. Dechlorination may affect
detection reliability and the chemical characteristics
of the water. Also, the bio-sentinels require a pro-
tected housing environment along with some sort of
nutritional supply to keep the sentinel organism alive
and healthy.
5.4 Establishing Monitoring
Requirements
Selection of the types, numbers, and locations of
monitors is dependent on the nature of the monitoring
program desired. These requirements depend upon
the overall monitoring objectives and the distribution
system site-specific requirements. For example, a
monitor used for regulatory purposes may need to
monitor different constituents than one used as part of
a process control or security system. Similarly, a
different monitor may be needed for a utility that uses
chlorine as the disinfectant compared to one that uses
chloramine. The site-specific monitoring require-
ments can be evaluated and represented in the
following terms:
• monitoring parameters,
• number and location of monitors,
• nonitor characteristics (e.g., detection limits,
sampling frequency, cost, false negatives/false
positives), and
• amenability to remote monitoring and SCADA
integration.
These requirements are further discussed in the
following subsections.
5.4.1 Monitoring Parameters
The parameters to be monitored depend strongly upon
the specific use of the monitor and upon utility-
specific situations. For regulatory purposes, the
regulations typically specify the minimum set of
parameters that must be sampled. For each system, the
regulating authority typically also specifies the
monitoring locations and frequency. A utility may
choose to analyze the water for additional parameters
and/or increase the frequency of monitoring in order
to address other water quality concerns.
For security monitoring, there are no regulations or
standards. Utilities can choose whether or not they
want to perform such monitoring and select the
parameters they will monitor. Generally, such
monitoring will be limited by budgets and by
technology. Research and development is being
conducted on security monitoring systems, in
conjunction with event detection platforms, that
measure standard parameters, such as TOC, pH,
turbidity, conductivity, chlorine, ORP and tempera-
ture. For both process control and security-related
monitoring, instrument response time is critical.
Therefore, online monitors are typically used in these
types of applications. The parameters monitored vary
widely depending upon the type of process and/or
security monitoring.
The goal of online monitoring for security purposes is
to automatically analyze the data to determine (1)
whether there is an indication of unusual contamina-
tion in the sample; and (2) what the likely contami-
nant is, based on the water quality signature of these
parameters.
5.4.2 Number and Location of Monitors
For selecting monitoring locations in distribution
systems, there are two related decisions: (1) how many
monitors to place in the system, and (2) where to
place them. The number of monitors is generally
controlled by the monitoring objective (e.g., regula-
tory requirement) or by budgetary factors, while the
location of monitors is a more complex issue that can
be addressed in many ways. For example, for compli-
ance with the TCR and the SWTR, there are specific
requirements as to the number of samples that must be
taken. For most other uses, the number of sampling
points (or the number of monitors installed) is
controlled by budgetary and financial constraints and
through comparison to the benefits associated with
the monitors. The following subsections summarize
an approach that can be used when the established
objectives do not clearly define the number and
location of monitors.
5-5
-------
A Reference Guide for Utilities
EPA's research in contamination warning systems (CWSs) at the T&E Facility is
developing data based on bench- and pilot-scale experiments that reveal how
traditional water quality parameters, if monitored online, can serve as triggers
for contamination events. Figure 5-1 shows the response of several instruments
to the injection of secondary wastewater into a DSS.
18:14 18:43
Time (Hours : Min)
Figure 5-1. Wastewater Injection: Free Chlorine and Associated Grab
Sample Results.
5.4.2.1 Number of Monitors
To select the optimum number of monitors for a
distribution system, theoretically one can perform a
simple cost-to-benefit analysis. If the overall life
cycle benefits of each monitor exceed its life cycle
costs, analysis would suggest that the monitor is
justified. Life cycle costs represent both the capital
and operational costs for the monitors. Depending
upon the location-specific requirements, as the
number of monitors increase, there may be economies
of scale or the unit cost may actually increase
disproportionately. The unit costs increase when the
additional monitors are placed in less convenient
locations where servicing and/or data communication
costs are higher. Frequently, budgetary constraints
may also limit the number of monitors that can be
QQ
o o
o o o
o
o
o
Figure 5-2.
Monitors.
Number of Monitors
Theoretical Example of Benefits from
deployed, even if benefits
justify their costs.
Figure 5-2 is a graphical
representation of benefits
associated with increasing the
number of monitors in a
distribution system. This
graph illustrates that typically
after a basic network of
monitors has been established
for a distribution system, the
incremental benefits gained
by installing additional
monitors follow the law of
diminishing returns. The
actual development of such a
graph is difficult because of
the need to explicitly quantify
benefits. In the case of water
security-related monitoring,
one could measure the value
based on population or
sensitive facilities (e.g.,
hospitals) protected by use of
online monitors. For other
types of monitoring situa-
tions, quantification of
benefits is more difficult.
Though a formal cost-benefit analysis may not be
feasible, this discussion provides a general framework
that can informally guide the design of a monitoring
network.
5.4.2.2 Optimal Monitor Locations
Historically, monitors/sensors have been placed in
distribution systems to meet regulatory requirements.
Their locations have been determined based on ease
of access and a general intuitive assessment of
representative locations. Lee et al. (1991) proposed a
method for locating monitors, based on the concept of
coverage, which is defined as the percentage of total
demand that is sampled by a set of monitors. Various
other researchers further addressed this issue using
alternative mathematical methods (Kessler et al.,
1998). Though widely cited, these methodologies
have rarely been applied in actual practice. However,
following the attacks of September 11, 2001, there has
been a renewed interest in the development of
monitoring technology and placement of monitors in
the distribution system as a mechanism for detecting
intentional contamination of distribution systems.
Many current studies are applying optimization
techniques to determine the optimal placement for
monitors in distribution systems based on a defined
objective function. Ostfeld (2004) and Ostfeld and
Salomons (2004) provide reviews of past work in this
area and present example mathematical formulations
5-6
-------
A Reference Guide for Utilities
A 17th century Italian economist, Vilfredo Pareto,
developed a method for comparing alternatives. Based
on his work, a situation is defined as being Pareto-
optimal if by reallocation you cannot make someone
better off without making someone else worse off. This
can be applied to evaluating monitors by examining the
diagram (Figure 5-3) where various monitoring options
are compared in terms of their cost and some measure of
effectiveness. Just looking at alternatives A and B, we
can say that A is better than B because it costs less and
is more effective. By comparing all potential alterna-
tives, we can define a Pareto front. All alternatives
located on that front are better than alternatives located
to the right and below the front. This provides a useful
conceptual mechanism for evaluating alternative
monitoring schemes. For additional information on the
work of Pareto, see Johansson (1991). For more details
on the application of Pareto's concepts in the area of
optimization related to water distribution system
analysis, see Walski et al., 2003).
Pareto front 2
3,
O
O O O
O
B
Cost
Figure 5-3.
Diagram.
Pareto-Optimal Cost Effectiveness
using genetic algorithm solution techniques. Their
methodology finds an optimal layout of an early
warning detection system comprised of a set of
monitoring stations aimed at capturing contamination
from external sources, nodes, or tanks under EPS
conditions. Berry et al. (2004) developed an optimi-
zation program that considers the maximum volume
of contaminated water exposure at a concentration
higher than a defined safe level. The method uses an
integer programming optimization technique to place
a limited number of "perfect" sensors in the pipes or
junctions of a water network so as to minimize the
expected amount of exposure to the public before
detection, assuming the attack occurs on a typical day.
Watson et al. (2004) use mixed-integer linear pro-
gramming models for sensor placement over a range
of design objectives. Using two case studies, they
Bahadur et al. (2003) describes an approach using
PipelineNet in which GIS data and hydraulic model
results are used to guide the manual placement of
monitors in order to fulfill some general criteria. In a
case study conducted with personnel at a water utility,
25 potential monitoring sites were identified and
subsequently reduced to two best sites using the GIS/
PipelineNet framework. This approach is more closely
related to the traditional methods for locating moni-
tors compared to the optimization techniques de-
scribed in this section.
showed that optimal solutions with respect to one
design objective (e.g., population exposed) are
typically highly sub-optimal with respect to other
design objectives (e.g., time for detection). The
implication is that robust algorithms for the sensor
placement problem must carefully and simultaneously
consider multiple, disparate design objectives.
In general, the optimization methods described above
are experimental approaches that have been applied
only to hypothetical or small water systems and are
based on assumptions about the availability of
monitoring technology, ability to define explicit
objective functions, and limited incorporation of the
variability of water system operation. Further
research and development is needed before this
technology is ready for routine use.
5.4.3 Monitor Characteristics
The following characteristics of monitors must be
evaluated prior to selecting an appropriate device:
• Minimum detection limit (MDL) - The
minimum detection limit is the lowest
concentration or value at which the monitor can
dependably detect the constituent of interest.
The MDL can vary for different constituents,
different technologies, or for different
implementations of the same technology and
constituent.
• False negatives/false positives - Two forms of
errors associated with a monitor are false
positives and false negatives. A false positive
exists when a monitor reports, incorrectly, that it
has detected a constituent where none exists in
reality. A false negative exists when a monitor
reports, incorrectly, that a constituent was not
detected when, in fact, it was present. False
positives can lead to unneeded responses, and
repeated false positives will lead to a lack of
confidence in the instrument. Lack of detection
associated with a false negative results in no
response to a real contamination event and can
expose consumers to contaminants in the
system.
5-7
-------
A Reference Guide for Utilities
The heightened level of concern over the need to
protect water distribution systems has led to the
initiation of research into the development of CWSs for
both source and finished waters (Clark et al., 2004a).
CWSs are intended to reliably identify low probability/
high impact contamination events in source or distrib-
uted water. The International Life Sciences Institute
(ILSI) developed a report (ILSI, 1999) focused on the
development of environmental warning systems (EWSs)
for source water. The same development principles
apply to distribution systems. EWSs applied to
distribution systems are commonly referred to as CWSs.
The following design requirements for EWSs were
identified by ILSI in their report:
• provides warning in sufficient time to respond to
a contamination event and prevent exposure of
the public to the contaminant,
• capable of detecting all potential contamination
threats,
• remotely operable,
• identifies the point at which the contaminant was
introduced,
• generates a low rate of false positive and false
negative results,
• provides continuous, year-round surveillance,
• produces results with acceptable accuracy and
precision,
• requires low skill and training, and
• be affordable to the majority of public water
systems.
A key aspect of an effective EWS will be the need for it
to operate in a remote monitoring and reporting mode.
• Sampling frequency - The rate at which a
monitor analyzes and reports a value is the
sampling frequency. This may vary from a few
seconds or less for an instrument such as a
pressure gage to an hour or more for instruments
that take longer periods to perform the analysis
such as a gas chromatograph. For grab
sampling, this delay may be even higher. Some
instruments can be set for different sampling
frequencies. More frequent sampling may result
in higher operating costs, shorter battery life,
increased data storage requirements, or
increased communication needs.
• Amenability to SCADA integration - The
monitor's ability to be online and integrated
into some sort of SCADA or remote data
acquisition system is critical if multiple remote
locations are monitored simultaneously. Most
current online monitors have analog (e.g., 4-20
mA, 1-20 V) or digital signal (e.g., RS232,
RS485) outputs that provide the ability to
remotely collect and store data at a central
location for analysis.
• Operation and maintenance requirements - The
operational requirements of monitors can vary
significantly and may strongly impact the
selection process. Issues include the electrical
needs, expendable material needs (e.g.,
reagents, wear related components), temperature
and humidity requirements, needs to handle
waste streams from the monitor, and other
factors related to the housing of the monitor.
Similarly, the maintenance requirements of the
monitors will also impact the selection process.
Issues such as how frequently a technician must
service the monitor in the field and the level of
expertise required to service the device are
important considerations when evaluating
monitors.
• Combinations of monitors - The ability of a
monitoring system to reliably detect a
contamination event generally increases with
multiple monitors working in tandem. For
example, a single monitor that reports a signal
slightly above the noise level may easily be
dismissed. However, if multiple monitors at
several locations in close proximity or several
instruments at the same location monitoring for
different parameters all detect a potential event,
a more forceful and rapid response is likely. An
ongoing area of research is the development of
data mining algorithms that can differentiate or
detect a signal above background levels that are
not normally observed in the monitored system.
• Costs - The cost of monitoring systems can
vary over several orders of magnitude. A single
simple instrument monitoring for a physical
parameter such as conductivity may cost less
than $1,000. The cost of a multi-parameter
physical monitor is typically in the
neighborhood of $10,000. More complex
instruments such as a TOC monitor or a GC/MS
cost in the range of $25,000 to $90,000. The
cost of more complex instruments or a
monitoring station containing multiple
instruments can easily exceed $100,000 in
capital cost. Installation and ongoing
maintenance costs are frequently site-specific
and vary according to environmental
conditions.
5.4.4 Amenability to Remote Monitoring and
SCADA Integration
For a comprehensive network-wide water quality
remote monitoring program, it is essential to ensure
that the system and its monitored components are
amenable to remote monitoring and SCADA Integra-
5-8
-------
A Reference Guide for Utilities
tion. The SCADA component adds the element of
control to the monitored network. Most utilities have
some sort of SCADA functionality to automate and
monitor the key water treatment and/or distribution
operations. The control logic is typically triggered
based on a specified time and/or event. For example,
the pumps may be set to fill a distribution system tank
at midnight and when the tank level monitor detects
that the tank is full (an event reported through the
SCADA system), the control logic to turn off the
pumps is initiated. This type of control logic can be
enhanced to perform control functions based on
detection of water quality change in the distribution
system. However, to achieve this functionality, one
needs to understand the following three major
components of a remote monitoring and/or control
system (orSCADA):
• online sampling instruments (e.g., pH, ORP)
and/or control devices (e.g., pump, valves),
• SCADA or remote monitoring network, and
• field wiring and communications media.
These components are discussed briefly in the
following subsections.
Electric power is generally required for operating these
components. If electric power is not readily available
at the desired location where a monitor is to be in-
stalled, consider the costs for installing a suitable
power apparatus (e.g., a solar panel, battery pack).
5.4.4.1 Online Sampling/Control Devices
Online sampling/control devices can be the most
expensive component of a SCADA system. The
sensors, switches, monitors, and controllers used in a
SCADA system may vary widely, depending upon the
parameters that need to be controlled and/or observed.
The cost for online sampling devices can range from a
few hundred dollars to over $100,000. Control units
such as sample feed pumps or shut-off valves are less
expensive (Panguluri et al., 1999). Costs associated
with maintenance and calibration of the online
sensors when planning the acquisition and implemen-
tation of a remote monitoring network should also be
considered.
5.4.4.2 SCADA or Remote Monitoring Network
Larger utilities typically use some type of SCADA
system for water distribution system control that can
easily be integrated to include online sampling
instrumentation in a cost-effective manner. Also,
recent advances in electronic hardware and software
technologies have resulted in several cost-effective
SCADA alternatives for smaller systems. A micropro-
Sensors and Transducers: A sensor responds to a
physical and/or chemical stimulus, such as thermal
energy, flow, light, chemical, pressure, magnetism, or
motion. A transducer takes the measured physical and/
or chemical phenomenon (e.g., pressure, temperature,
humidity, and flow) and converts it to an electrical
signal. In each case, the electrical signals produced are
proportional to a physical and/or chemical quantity
being measured based on a pre-defined relationship.
The electrical signals generated by transducers often
require "conditioning." Depending upon the trans-
ducer, a signal conditioner can be used to perform one
or more conditioning functions, such as noise filtration,
amplification, linearization, isolation, and excitation.
cessor-based "smart" SCADA system can be used in
remote locations by small system operators where
direct online communication is expensive. Smart
systems have higher initial costs, but overall costs are
reduced since the communication costs (e.g. long-
distance phone costs) are negligible because most of
the burden is transferred from the main computer to
the individual SCADA unit at the remote site
(Panguluri etal., 1999). Newer SCADA units are
fairly inexpensive, with capital costs ranging between
$500 (PC card-based units and remote data collection
nodes) and $5,000 (independent PC-based full
SCADA units).
The data acquisition hardware processes the digital
and analog inputs/outputs from various online
sampling and control devices. For monitoring
systems, the hardware typically processes the analog
data measured from various instruments and transfers
it to a computer system for display, storage, and
analysis. In a monitoring/control system (SCADA)
scenario, the hardware would process both analog and
digital inputs (typically from a field device) and
outputs (to perform control functionality). The
application software provides the operator the
display, control, and analysis (trends and reports) of
collected data.
5.4.4.3 Field Wiring and Communication Media
Depending upon availability, cost, user preference,
and the relative location of the sensors to the data
acquisition system, the communication media can be
either wired (e.g., direct, phone line) or wireless (e.g.,
radio, cellular). In field environments, distributed
input/output (I/O) is typically employed. A remote
data acquisition hardware unit employed at the field
location performs the appropriate signal conditioning
and transmits the data to a central hub. More recently,
mesh or grid computing systems are used in remote
locations to add redundancy in cases of link failures.
The field wiring between the sensor and the remote
data acquisition hardware unit is usually direct
wire.
5-9
-------
A Reference Guide for Utilities
Depending upon the area covered and availability, in
some cases it may be preferable to use some form of
radio communication devices. The available radio
communication devices operate mostly in the very
high frequency (VHP) or ultra high frequency (UHF)
range. The VHP frequencies range between 30-300
megahertz (MHZ) and the UHF frequencies range
between 300-1,000 MHZ. In U.S., most of the
available VHF/UHF radio frequencies are licensed.
The unlicensed bands available include the industrial,
scientific and medical device channels with frequency
ranges between 902 - 928 and 2,400 - 2,484 MHZ.
The unlicensed bands do not have any connection or
monthly fee requirements.
Typically, direct wire and phone line (including
cellular) communication media are inexpensive. The
primary limitations associated with selecting the
communication media include installation and
operating costs, which can vary between $200 (for a
simple telephone or cellular modem) and several
hundred dollars for a satellite-based system per
location. Ongoing monthly operating costs can range
from $25 for a phone line to approximately $200 per
month for satellite-based services within the U. S (per
monitored location).
5.5 Engineering and Evaluating a
Remote Monitoring System
Once all of the basic requirements have been estab-
lished (e.g., objectives, parameters, location) as
outlined in the previous section (Section 5.4) and the
requirements indicate a need for a system-wide remote
monitoring program for water quality, the following
additional site-specific needs should be evaluated for
water quality monitoring in a distribution system
(Pangulurietal., 1999):
• What are the complexities of the distribution
system (size, location)?
• What locations are best suited for sampling and/
or control system installation?
• Is sufficient flow and water pressure available
for online instruments?
• Is there an existing SCADA system available?
• What types of communication media are
available at the selected locations?
• How many parameters are going to be
monitored and/or controlled at each location?
• What other site-specific information (e.g.,
availability of power, access, security) will be
needed?
Additonal factors to be considered are (Haught and
Panguluri 1998):
• system features (e.g., ease of operation,
customization, networkability, operator
security),
• cost (initial, training, service agreements, and
operation and maintenance), and
• vendor support (hardware and software
upgrades and remote diagnosis).
It is important that each site is evaluated individually
for appropriate SCADA system selection. The cost of
SCADA software has plummeted over the past few
years. For example, the cost of one commercially
available graphical (Windows-based) SCADA software
package has dropped from $30,000 in the early 1990s
to $2,000 today.
Prior to selecting and implementing a remote monitor-
ing network, one should evaluate the options care-
fully. Engineering a remote monitoring system is a
difficult task that typically involves many factors:
multi-dimensional objectives, changing needs, rapid
Besides the aforementioned immediate needs (e.g., ease
of operation, customization, networkability), SCADA
system features include:
• Scalability: This allows for future growth with
respect to addition of I/O blocks with more
channels or advance capabilities. These I/O
channels are used to communicate with various
field monitoring instruments (sensors) and control
devices.
• Local Memory: The SCADA hardware must also
contain sufficient local memory to store the
monitored data for extended periods of time in
case of communication failures.
• Remote operation and diagnosis: In the event of
brownouts or blackouts, the field SCADA units
should normally self-boot upon resumption of
power supply. The field SCADA units should also
allow for remote diagnosis.
• Call-out feature: This feature allows the system's
software to notify appropriate personnel if
problems develop with a treatment system or
water quality. This feature can greatly enhance
operator response in emergency situations and
prevent costly shutdowns and loss of water and/or
water quality.
• Open Database Connectivity (ODBC): This
feature allows for open communication with other
databases and tools that can be integrated to
provide additional features. The data then can
also be used for network modeling.
5-10
-------
A Reference Guide for Utilities
technological change, conflicting technical claims,
and budgetary constraints. The following subsection
presents general methods for evaluating and assessing
alternatives followed by a set of specific criteria for
evaluating alternative monitoring systems.
5.5.1 Remote Monitoring System Evaluation
In order to justify a remote monitoring system and to
select the best monitoring system, it would be ideal if
one could evaluate the benefits derived from monitor-
ing and compare them to costs and choose the system
that maximizes net benefits subject to budgetary
constraints. Depending on the uses of the monitoring
data, monitoring benefits may be associated with:
• reduced risks from an intentional or accidental
contamination event,
• improved understanding of the variation in
water quality of a distribution system,
• enhanced operation if the data are used as part
of a process control system, and
• increased compliance if the information is used
for regulatory purposes.
5.6 Monitoring Case Studies
EPA has conducted research into the use of remote
monitoring and control technology alternatives for
many years. These projects have involved both water
treatment systems and water distribution (Clark et al.,
2004b). The agency's first research project that
incorporated real-time monitoring at a remote
location was conducted at the T&E Facility. The
initial research was focused on evaluating SCADA
systems for small drinking water package plants. The
goal was to demonstrate that SCADA systems could
be used to monitor and control several small plants
remotely from a centralized location at one time
(Haught and Panguluri, 1998). The following case
studies represent some of the highlights of the
research and collaboration with different water
utilities.
5.6.1 Rural Community Application
In May 1991, EPA provided funding to support a
research project titled "Alternative Low Mainte-
nance Technologies for Small Water Systems in
Rural Communities" (Goodrich et al., 1993). This
project involved the installation of a small drinking
water treatment package plant in a rural location in
West Virginia. The primary objective of this study
was to evaluate the cost-effectiveness of package
plant technology in removing and disinfecting
microbiological contaminants. The secondary
objectives of this project included: remote monitor-
ing and automation of the system to minimize the
O&M costs, assessment of the community's accep-
tance of such a system, ability to pay, and the effect
of the distribution system on water quality at the
tap. The following is a brief summary of the overall
project.
The treatment system was located in rural Coalwood
(McDowell County), WV, approximately 12 miles
from the McDowell County Public Services Division
office. Prior to 1994, an aerator combined with a slow
sand filter was being used for water treatment at this
site. This combined unit had been operational for
over 30 years and needed substantial repairs. The
water flowed by gravity from an abandoned coal mine
to an aerator built over a six-foot-diameter slow sand
filter. A hypochlorinator provided disinfection of the
treated water, and the water flowed by gravity through
the distribution system to the consumer. The volume
of water from the mine was considered sufficient for
the small rural community.
Based on a review of existing technology, EPA
determined that a packaged ultrafiltration (UF) system
would be ideally suited for this location. In 1992, a
UF unit was purchased and installed at this site. In
1996, EPA developed, installed, and tested a remote
monitoring system at the site. The system used
commercially available hardware along with EPA-
developed software. The software was not user-
friendly and the overall cost of ownership was very
high. Therefore, in 1998, EPA updated the SCADA
system with a scalable commercially available off-the-
shelf, user-friendly SCADA system. The total cost
(including instrumentation, technical support,
training, and set-up) was approximately $33,000.
EPA installed similar SCADA systems at Bartley and
Berwind sites in McDowell County, WV, for remote
monitoring of water quality.
5.6.2 Washington D.C. Remote Monitoring
Network
Following a number of coliform violations, EPA's
Region 3 office directed the Washington D.C. Water
and Sewer Authority (WASA) to implement a number
of corrective actions for its water distribution system
(Clark et al., 1999). Remote monitoring of water
quality parameters within the distribution system was
identified as being one possible method for identify-
ing water quality problems. Consequently in 1997,
EPA initiated a study to install a remote network at
various locations in Washington D.C. to monitor
water quality within the distribution system (Meckes
et al., 1998). The WASA staff teamed with EPA to
select appropriate locations within the distribution
system for installation of online sampling stations.
Following are some of the study objectives:
• development of methods to monitor real-time
water quality at various locations within the
5-11
-------
A Reference Guide for Utilities
distribution system,
• field evaluation of sensors and remote
monitoring technologies for inclusion in the
network,
• development of effective methods to publish
real-time data that enhanced consumer
confidence,
• evaluation of costs associated with
implementing such systems, and
• identification of potential problems and
suggestions for remedial actions when
implementing remote monitoring networks.
Free chlorine, pH, temperature, and turbidity were
selected as the monitored parameters based on the
availability of online sensor technologies. The
selection was based on the premise that these were
parameters which could be reliably monitored
continuously and the selected instruments required
limited maintenance. Additionally, WASA used their
SCADA system to track various operating parameters
within the distribution system. During the evalua-
tion, it was clear that use of the existing SCADA
system to manage the monitored data provided clear
advantages over other available systems. Using the
existing SCADA system minimized long-term on-site
support costs.
After suitable location(s) were identified, customized
sampling and monitoring systems were built. The
remote monitoring system in Washington D.C. was
implemented in three phases. In the first phase, a
remote monitoring system was installed at the Fort
Reno #2 tank (Figure 5-4), which provided security
and easy access to the distribution system. Subse-
quently, based on initial success at this location, two
other sites (Bryant Street and Blue Plains) were
selected and added
to the remote
monitoring network
in the second phase.
The third phase
involved the
development of a
Web-based applica-
tion to publish the
real-time data in
order to enhance
consumer confi-
dence.
Figure 5-4. Fort Reno #2
Remote Sampling System.
though WASA's SCADA system used a proprietary
operating system, it provided a personal computer
(PC) link which was used to dump data into a regular
PC for further processing. The hardware-based feature
enabled tight security; an authorized end user could
only copy the relevant data published on the PC and
could not directly access the SCADA system. This
feature also eliminated any potential interference
between the sampling system data and other distribu-
tion system operations data. Unfortunately, the EPA
funding for this study was terminated and, as a result,
the systems and the Website are currently not opera-
tional. The overall project, however, did demonstrate
that such systems could be developed and operated.
Figure 5-6 shows some of the output data for the Fort
Reno tank which indicates the loss of disinfectant
chlorine levels at night. Clearly, this type of informa-
tion can be used to improve system operations to
better maintain the water quality.
5.6.3 Tucson Water Monitoring Network
Based on a grant received from the EPA's Environmen-
tal Monitoring for Public Access and Community
Figure 5-5 shows the
relationship between
the SCADA system
and the transmission
of the data. Al-
Figure 5-5. WASA Remote Monitoring System Layout and Data
Transmission Scheme.
5-12
-------
A Reference Guide for Utilities
mmmmmm&mmm
Figure 5-6. Monitoring Data for Fort Reno Tank.
Tracking (EMPACT) Program, the city of Tucson
implemented a comprehensive water quality monitor-
ing program. The city's EMPACT goals included the
following: implementing enhanced monitoring of the
utility's potable distribution system, providing the
community with near real-time water quality informa-
tion on Tucson Water's Website
(www.cityoftucson.org/water), and creating commu-
nity partnerships to better inform water consumers
about water quality and resource issues. The water
quality monitoring and data collection tools provided
through EMPACT also enables the utility to track and
respond to real-time changes in system water quality.
Tucson Water's distribution system consists of one
central drinking water distribution system that serves
the majority of the customers and ten isolated
drinking water distribution systems. All eleven
drinking water distribution systems cover a service
area of 300 square miles and serve 680,000 customers
in the Tucson metropolitan area. The two types of
source water that supply the central distribution
system are native groundwater and renewable
recharged surface water from the Colorado River. The
source water that supplies the ten isolated distribution
systems is groundwater.
For the purposes of monitoring, the central distribu-
tion system is divided into ten water quality zones
and each isolated distribution system is considered an
individual water quality zone. Figure 5-7 shows the
zone map. A water quality zone is defined as an area
of the distribution system that is similar in water
quality characteristics, water pressure, geographical,
and political boundaries. Each water quality zone has
a set number of dedicated sampling stations and
points-of-entry (POE). The dedicated sampling
stations monitor the quality of the drinking water in
the distribution system before delivery to the cus-
tomer. The POEs are usually individual wells that
represent the water quality of a single well or in a few
cases, combined POE systems that represent the
collective blended water quality from a group of wells
that directly supply Tucson's drinking water.
In total, there are 262 dedicated sampling stations and
approximately 154 active POEs located within the
multiple distribution systems. In addition, 22 online
water quality stations (for monitoring: chlorine
residual, total dissolved solids, pH, and temperature)
are located throughout the central distribution system
at strategic locations, such as reservoirs, well sites,
and booster stations, as one of the primary objectives
of the EMPACT program.
Figure 5-7. City of Tucson Water Quality Zone Map.
Figure 5-8 depicts a continuous water quality
monitoring station. The monitoring frequency ranges
from tri-annually (for grab sample locations) to every
60 seconds (for continuous monitoring stations),
depending on the location and specific monitoring
program that is being utilized for that location.
The comprehensive water quality monitoring program
encompasses the entire distribution system. Source
waters are monitored and sampled according to the
Arizona Department of Water Resources and the
Figure 5-8. Continuous Water Quality Monitoring Station.
5-13
-------
A Reference Guide for Utilities
Arizona Department of Environmental Quality
(ADEQ) regulations, while the drinking water is
monitored according to EPA and ADEQ regulations
and drinking water standards. Drinking water is also
evaluated against a set of consumer-established water
quality goals. Special-purpose samples are taken to
characterize and track changing trends in water
quality for both source water and drinking water. All
data sets are utilized to track and monitor changes in
water quality to learn the baseline water quality
operating parameter levels and also to be able to
identify and react appropriately when a contamina-
tion event occurs. Most of the analysis is conducted
by the utility's water quality laboratory and all the
results are tracked through the Water Quality Manage-
ment Division.
All 262 dedicated sampling locations are monitored
at least once each month for total coliform and
chlorine residual, while 26 other parameters are
monitored once every three months. Based on the
water quality measurements collected each month
from these 262 sampling locations, the trends in water
quality conditions are determined for each water
quality zone and for the distribution system as a
whole. This information can be found on the afore-
mentioned Web site in the Water Quality section
under Tucson's Water Quality and Water Quality in
My Neighborhood links. The water quality informa-
tion displayed on two interactive maps shows data
charts and tables for each location that is sampled
under the Water Quality program. In addition, the
information provided to all Tucson water customers in
the annual water quality report or consumer confi-
dence report is based on POE monitoring data.
5.7 Summary and Conclusions
Distribution system monitoring is intended to identify
the spatial and temporal variations in water quality
that take place in a drinking water system. Monitor-
ing data can be used to satisfy various objectives,
such as regulatory requirements, security require-
ments, or process control requirements. The costs of
implementing such a system can best be justified if
the resulting data can be used for more than one of the
aforementioned objectives.
A monitoring program can implement either routine
grab sampling or continuous monitoring. A com-
bined approach, utilizing both continuous and grab
sampling data, may prove to be very effective as the
basis for a comprehensive system-wide monitoring
plan. In the past, distribution system online monitors
were typically housed in a controlled environment
with sample lines from the distribution system to the
instrument. This resulted in most instrumentation
being located at facilities such as tanks and pump
stations. The instrumentation was sometimes con-
nected to a SCADA system, so that results could be
communicated to a central office. Recently, some
instrumentation has been designed for installation in
manholes or for direct insertion into water distribu-
tion system pipes.
Vulnerability assessments performed by utilities and
various research studies have identified that water
distribution systems are vulnerable to intentional or
accidental contamination. In addition to hardening
systems to make it more difficult to contaminate a
system, monitoring as part of a CWS has emerged as a
logical approach to cope with potential contamina-
tion events. Monitors can also be used in a distribu-
tion system to provide real-time or near real-time
information on water quality. The data can then be
used to control treatment processes at a treatment plant
or in the water distribution system. However, this type of
program may not be practical for small systems.
SCADA is widely used in industrial environments and
by larger water utilities to control and monitor their
individual facility operations. However, water
utilities typically do not use available SCADA
systems for conventional water quality monitoring.
Water utilities typically monitor water quality
parameters by performing grab sampling on a sched-
uled or random basis that provides a periodic snap-
shot of the overall system. Current drinking water
regulations require all public water systems to
implement water quality monitoring for total coliform
to ensure that good quality water is provided to
consumers (EPA, 1996). Since the regulations do not
clearly specify that real-time monitoring of water
quality is required, utilities have been reluctant to
install and operate such devices.
After the events of 9/11, utilities have become more
interested in the potential for continuous water
quality monitoring. SCADA systems can assist in this
function by constantly monitoring water quality
within drinking water distribution systems. These
systems can potentially reduce the risk of security
related threats or even non-security related threats,
and detect undesirable water quality changes within a
system (Meckes et al, 1998).
Users should evaluate monitoring data appropriately for
errors and inconsistencies before commencing actions
based on acquired data. Each component in a monitor-
ing system is a potential source of error. For example, a
remote monitoring system could have data errors for one
or more of the following reasons: instrument errors and
spikes, SCADA data errors related to system failure,
backfilling due to communication failure, timing errors,
or missing data. It is important to validate data and
understand routine changes in water quality due to
system-specific operations.
5-14
-------
A Reference Guide for Utilities
Monitoring equipment should be chosen appropri-
ately after establishing the monitoring requirements.
The individual monitor characteristics, costs, and
amenability to SCADA integration are key to effective
implementation. Each system should be individually
examined and engineered for implementation.
The monitoring case studies presented in this chapter
demonstrate the manner in which effective monitoring
systems can be implemented in small, medium, and
large distribution systems. If the data are used for
responding to a contamination threat, it is important
to understand the movement of water in the system.
References
ASCE. Interim Voluntary Guidelines for Designing
an Online Contaminant Monitoring System. Pub-
lished by American Society of Civil Engineers,
Reston,VA. 2004.
Bahadur, R., W. Samuels, W. M. Grayman, D. Amstutz,
and J. Pickus. "PipelineNet: A Model for Monitoring
Introduced Contaminants in a Distribution System."
Proceedings, ASCE-EWRI World Water & Environ-
mental Resources Congress, Philadelphia, PA. 2003.
Berry, J., W. Hart, C. Phillips, and J. Uber. "A General
Integer-Programming-Based Framework for Sensor
Placement in Municipal Water Networks." Proceed-
ings, ASCE-EWRI World Water & Environmental
Resources Congress, Salt Lake City, UT. 2004.
Clark R.M., WM. Grayman, S.G. Buchberger, Y. Lee,
and D.J. Hartman. "Drinking Water Distribution
Systems: An Overview" in Water Supply Systems
Security. Edited by L. W. Mays, McGraw-Hill, NY, pp
4.1-4.49. 2004a.
Clark, R.M., S. Panguluri, and R.C. Haught. "Remote
Monitoring and Network Models: Their Potential For
Protecting U.S. Water Supplies," in Water Supply
Systems Security, edited by L. W. Mays, McGraw-Hill,
NY, pp 14.1-14.22. 2004b.
Clark, R.M., G.S. Rizzo, J.A. Belknap, C. Cochrane.
"Water Quality and the Replacement and Repair of
Drinking Water Infrastructure: The Washington, DC
Case Study." Journal of Water Supply Research and
Technology - Aqua, 48(3): 106-114. 1999.
EPA. Drinking Water Regulations and Health
Advisories. Office of Water, EPA 822-B-B-96-002.
October, 1996
EPA. "Response Protocol Toolbox." Overview and
Modules 1 through 6. Can be downloaded at: http://
cfpub.epa.gov/safewater/watersecurity/
home.cfm?program_id=8#response_toolbox. 2003-
2004.
Goodrich, J., J. Adams, and B. Lykins, Jr. "Ultrafiltra-
tion Membrane Application for Small System." EPA
National Risk Management Research Laboratory.
1993.
Haught, R. C., and S. Panguluri. "Selection and
Management of Remote Telemetry Systems for
Monitoring and Operation of Small Drinking Water
Treatment Plants." Proceedings, First International
Symposium on Safe Drinking Water in Small Systems,
May 10-13, 1998, Washington, D.C. USA. 1998.
Hargesheimer, E., 0. Conio, and J. Popovicova
(Editors). Online Monitoring for Drinking Water
Utilities. AwwaRF - CRS ProAqua. 2002.
International Life Sciences Institute (ILSI). Early
Warning Monitoring to Detect Hazardous Events in
Water Supplies. ILSI Press, Washington D.C. 1999.
Johansson, P. An Introduction to Modern Welfare
Economics. Cambridge University Press, Cambridge,
UK. 1991.
Kessler, A., A. Ostfeld, and G. Sinai. "Detecting
Accidental Contaminations in Municipal Water
Networks." Journal of Water Resources Planning and
Management, ASCE. 124(4):192-198. 1998.
Lee, B., R. Deininger, and R. Clark. "Locating
Monitoring Stations in Water Distribution Systems."
Journal of AWWA, 83(7):60-66. July 1991.
Meckes, M. C., J. S. Mattingly, G. J. Papadopoulos, M.
Dosani, and S. Panguluri. "Real Time Water Quality
Monitoring of a Water Distribution System Using
Remote Telemetry." Proceedings, AWWA Distribution
System Symposium, Austin, Texas. September 20-22,
1998.
Ostfeld, A. "Optimal Monitoring Stations Allocations
for Water Distribution System Security" in Water
Supply Systems Security, edited by L. W. Mays,
McGraw-Hill, NY, pp 16.1-16.15. 2004.
5-15
-------
A Reference Guide for Utilities
Ostfeld, A., and E. Salomons. "Optimal Layout of
Early Warning Detection Stations for Water Distribu-
tion Systems Security." Journal of Water Planning
and Resources Management, ASCE. 130(5) :377-385.
2004.
Panguluri, S., R. C. Haught, M. C. Meckes, and M.
Dosani, 1999. "Remote water quality monitoring of
drinking water treatment systems." Proceedings,
AWWA Water Quality Technology Conference,
Denver, CO. November 1999.
Uber, J., F. Shang, M. Ploycarpou, and Z. Wang.
"Feedback Control of Booster Chlorination Systems."
AwwaRF, Denver, CO. 2003.
Walski, T.M., D.V. Chase, D.A. Savic, W Grayman, S.
Beckwith, and E. Koelle. Advanced Water Distribu-
tion Modeling and Management. Haestad Press,
Waterbury, CT. 2003.
Watson, J., H. J. Greenberg, W. E. Hart. "A Multiple-
Objective Analysis of Sensor Placement Optimization
in Water Networks." Proceedings, ASCE-EWRI World
Water & Environmental Resources Congress, Salt
Lake City, UT. 2004.
5-16
-------
A Reference Guide for Utilities
Chapter 6
Geospatial Technology for
Water Distribution Systems
Section 6.1 provides a brief summary of the history and development of geospatial data management
based on information extracted from many sources. This history is included in order to provide a context
for the current geospatial data management methodologies in use today by utilities. Readers who are
already familiar with the history may choose to skip Section 6.1 of this chapter.
Geospatial data identifies the geographic location and
characteristics of natural or constructed features and
boundaries on the earth. This information may be
derived from various sources of data, including
remote sensing, mapping, and surveying technolo-
gies. More simply, geospatial data is any information
in or on the earth that has a "where" component. This
can be a house address, a street intersection on a map,
a pump station with a coordinate location stored in a
facilities list, or the location of the sampling tap on a
diagram of a pump station. Thus, every object has a
geospatial data component based on its location.
Geospatial data provides a mechanism for incorporat-
ing geographic locations of various functions and
facilities in a distribution systems analysis. The cost
of incorporating map data into the water distribution
systems discipline is decreasing, which enables a
wider audience of users to perform powerful spatial
analyses over time, such as master plan development,
pipe break analysis, and locational information on
sensitive subpopulations (e.g., nursing homes,
schools). As these tools and datasets become more
commonly used and shared among engineers, new
efficiencies will be realized that will have a positive
impact on water distribution system management.
Water systems are by nature quite geographically
extensive and the location of a particular component
or feature may significantly affect its performance.
Source watersheds can cover hundreds or thousands of
square miles. Similarly, distribution systems can
cover vast areas. The operation of a water system
entails moving water from one location to another.
Elevation (Z), the third dimension of location (along
with the X and Y dimensions of a Cartesian coordi-
nate system), is an important factor in designing and
operating a water system. This illustrates that the
management of a water system is inherently a
geospatial issue.
Because of the spatial nature of water systems, many
aspects of managing a water system consist of using,
managing, and displaying geospatial data. This has
led to a variety of mechanisms ranging from maps and
plans to sophisticated, computerized database
management systems. The following is a list of some
of the computerized data systems that water utilities
typically use for managing their spatial data. These
systems will be discussed in greater detail later in this
chapter.
• GIS - Geographic Information System.
• CADD - Computer-Aided Design and Drafting.
• AM/FM - Automated Mapping (or Asset
Management)/Facilities Management.
• CIS - Customer Information System.
• DEM - Digital Elevation Model.
• GPS - Global Positioning System.
• SCADA - Supervisory Control And Data
Acquisition.
• LIMS - Laboratory Information Management
System.
• LIS - Land Information System.
• RDBMS - Relational Database Management
System.
• SDMS - Spatial Data Management System.
6-1
-------
A Reference Guide for Utilities
6.1 History of Geospatial Data
Management
Geospatial data is of interest in many professional
fields and each of these fields has approached the
issue of using and managing geospatial data in a
different manner. Examples include the following:
• Cartographers concentrate on making maps.
• Surveyors emphasize accurate capture of
locational information on natural and manmade
land features.
• Engineers use spatial data to draw construction
plans and, more recently, use it as input for
various types of models.
• Planners use maps and spatial data to assess
growth and to determine the suitability of land
to support a particular type of development.
• People in the public works area are concerned
with managing assets such as streets, sewers, and
water lines, which all have a spatial component.
• Fields such as the military, engineering, mining,
and hydrology are interested in topographic
(elevation) data.
As a result of these varying interests in "spatially
arrayed data," the tools and methodologies for
managing these data have evolved from many
directions and recently there has been a significant
move towards integrating the basic concepts. A brief
history of geospatial data management is presented
below, organized by the various disciplines that have
influenced this field. The needs of the water industry
in the spatial arena cross each of these areas. Geo-
spatial data management in the water industry will be
discussed in greater detail in a later section of this
chapter.
6.1.1 Mapping, Surveying, and Remote Sensing
Mapping is the oldest of the geospatial disciplines.
Examples of maps date back many millenniums. A
wall painting, dating back to around 6200 B.C. in
Turkey, depicts the positions of the streets and houses
of the town together with surrounding features such as
the volcano close to the town. The Babylonians
produced clay tablets containing maps that date back
to around 1000 B.C. Other early maps were prepared
by the Egyptians and Chinese (O'Connor and
Robertson, undated).
Many of the advances in map making are attributed to
the Greeks. Around 350 B.C., Aristotle argued that
the earth was a sphere and around 250 B.C., Era-
tosthenes accurately calculated the circumference of
the earth. In 140 A.D., Ptolemy's eight-volume Guide
to Geography was written and provided the basic
principles of cartography. It introduced the concept of
map projections and attempted to map the known
world, giving coordinates of the major places in a
system akin to the present day latitude-longitude
system. This document served as the definitive
reference on geography for over a thousand years and
was later translated into Latin and printed in 1475.
The 16th century saw the introduction of globes and
many improvements in the mathematical basis of
cartography. Gerardus Mercator developed a wall
map of the world in 1569 on 18 separate sheets (see
Figure 6-1). In the "Mercator projection," lines of
longitude, lines of latitude, and rhomb lines all
appear as straight lines on the map. This projection
was a great aid to navigators and is still in use today.
With the basis of cartography well in hand almost 500
years ago, the cartographic methodologies continued
to evolve. Additionally, emphasis was placed on
methods of accurately establishing the coordinates of
places of interest. This led to the field of surveying
and, more recently, the field of remote sensing.
Figure 6-1. Mercator's Map of the World in 1569
(Whitfield, 1994).
"Surveying is the science and art of measuring
distances and angles on or near the surface of the
earth. It is an orderly process of acquiring data
relating to the physical characteristics of the earth and
in particular the relative position of points and the
magnitude of areas. Evidence of surveying and
recorded information exists from five thousand years
ago in places such as China, India, Babylon, and
Egypt" (Queensland Government, undated). Some
key inventions in the area of surveying include the
following:
• Knotted ropes - Measuring device developed
by the Egyptians and used in construction of
the pyramids.
• Levels - Mechanism developed by the
Egyptians composed of a hanging "plumb bob"
used to establish a level surface.
6-2
-------
A Reference Guide for Utilities
Magnetic compass - Device for determining
north direction. Developed by Chinese circa
200 B.C. and composed of a magnetic
"lodestone." Later used by Chinese for
navigation.
Theodolite - An instrument, graduated in 360
degrees, used in the mid-1500s by an
Englishman, Leonard Digges, to measure
angles.
Alidade and Plane table - Sighting mechanism
and flat table developed in 1590 for mapping
the surface features of the earth. Attributed to
Jean Praetorius.
Quadrant and Sextant - The quadrant is an
apparatus developed in 1730 by John Hadley
for measuring angles of celestial bodies. This
led to the development of the sextant, which is
a precision instrument made from brass or
aluminum that is used for ocean navigation
where celestial observations are taken to plot a
ship's position.
Transit - The transit is used to measure vertical
and horizontal angles and may also be used for
leveling; its chief elements are a telescope that
can be rotated (transited) about a horizontal and
about a vertical axis, spirit levels, and
graduated circles supplemented by Vernier
scales. Attributed to W.J. Young in Philadelphia
in 1831.
Electronic Distance Measurement (EDM) -
Starting in the 1950s, electronic distance
measuring instruments were developed and
have now largely replaced traditional methods
for measuring distance. Horizontal distances
are measured using a variety of instruments that
employ a laser beam aimed at a reflector station.
Low-cost instruments that employ sound waves
or infrared beams are also available.
GPS - The use of GPS in surveying procedures is
the most recent and revolutionary change to
influence land measurement. GPS was designed
and built and is operated and maintained by the
U.S. Department of Defense. Originally called
the Navstar GPS, it was first brainstormed at the
Pentagon in 1973. In 1978, the first operational
GPS satellite was launched; by the mid-1990s,
the system was fully operational with 24
satellites. The basic principle behind GPS is the
measurement of distance between satellites and
the receiver. The distance to at least 3 satellites
must be known in order to find out a position.
Satellites and receivers generate duplicate radio
signals at exactly the same time. As satellite
signals travel at the speed of light (186,000
miles per second), they only take a few
hundredths of a second to reach the GPS
receiver. This difference and the speed at which
the signal travels is used in the equation to find
out the distance between the GPS receiver and
the satellite (Radio Shack, 2004). GPS is now
also being used to provide information on
elevations.
Local governments frequently store survey informa-
tion on parcels in an LIS. This information can
include property ownership, construction date, land
assessment, and land taxation. This information may
be linked to a computerized database system for
storing the geographic coordinates of the parcels.
Remote sensing refers to imagery from airplanes or
satellites. Some early examples of remote sensing
include: aerial photography from a balloon in 1859
by Gaspard Felix Tournachon in an attempt to
conduct a land survey; use of light cameras attached
to pigeons in Bavaria in 1903 to monitor troop
positions; photographs of San Francisco following the
1906 earthquake by George Lawrence from a kite; and
the work of a photographer who accompanied Wilbur
Wright on one of his first demonstration flights in
1909. More serious aerial photography was conducted
during World War I and II and during the Cold War
period.
Aerial photography has become a staple item in the
development of maps and documentation of land use
changes. The National Aerial Photography Program
(NAPP) is an interagency Federal effort coordinated
by the USGS, which uses NAPP products to revise
maps. Other agencies have varied uses for these
photographs, which are taken on a 5- to 7-year cycle
and produced to rigorous specifications. The NAPP
effort encompasses the entire lower 48 states and
Hawaii. The photos are acquired from airplanes flying
at an altitude of 20,000 feet using a 6-inch focal
length camera resulting in a scale of 1:40,000. Each
9-inch by 9-inch photo (without enlargement) covers
an area of slightly more than 5 miles on a side. The
NAPP effort began in 1987 and replaced the National
High Altitude Photography (NHAP) program which
was initiated in 1980. Strict specifications regarding
sun angle, cloud cover, minimal haze, stereoscopic
coverage, and image inspection were followed and all
NAPP photography is cloud-free (USGS, undated).
Satellite remote sensing can be traced to the early
days of the space age (National Aeronautics and
Space Administration [NASA], undated). On April 1,
1960, the Television and Infrared Observation
Satellite (TIROS 1) was launched, which proved that
satellites could observe Earth's weather patterns. In
1966, the Environmental Science Services Adminis-
6-3
-------
A Reference Guide for Utilities
tration (ESSA) Satellites I and II gave the United
States its first global weather satellite system. In 1972,
NASA began the Landsat series with the launch of the
Earth Resources Technology Satellite 1, which was
later renamed Landsat 1 by NASA. Figure 6-2
illustrates an image derived from Landsat. Subse-
quently, U.S. governmental satellites, such as Landsat
7, are still gathering consistently calibrated imagery
of the earth under the Earth Observation Satellite
(EOSAT) program. Satellites of other governments
(SPOT-France) and private satellites (GE, Digital
Globe) have expanded the routine availability of
imagery, and further enhanced the resolution of the
data collected.
Figure 6-2. Landsat Thematic Mapper™ Images of
the Missouri River Floodplain Near Glasgow,
Missouri. (USGS, 1993).
6.1.2CADD
Over the past quarter of a century, CADD has revolu-
tionized the way in which engineers and architects
perform their work. The basis for CADD was laid by
Ivan Sutherland's 1963 Ph.D. thesis at Massachusetts
Institute of Technology (MIT) titled, "Sketchpad: A
Man-machine Graphical Communications System"
(Sutherland, 2003). Sutherland used a lightpen to
create engineering drawings directly on the Cathode
Ray Tube (CRT). His thesis laid out virtually all of
the graphical human interface issues. Sketchpad
pioneered the concepts of graphical computing,
The acronyms CAD, CADD, CAM, and CAE refer to
"computer aided" methodologies used in various fields
of engineering. CAD can stand for "computer aided
drafting" or "computer aided design". CADD can
mean "computer aided design and drafting" or "com-
puter aided drafting and design". CAE refers to
computer aided engineering and CAM refers to
"computer aided manufacturing". The fields of CAD,
CAM and CAE overlap and are frequently lumped into
a single field of CAD/CAM/CAE.
including memory structures to store objects, rubber-
banding of lines, the ability to zoom in and out on the
display, and the ability to make perfect lines, corners,
and joints. This was the first GUI long before the term
was coined.
In the late 1960s and early 1970s, several companies
were founded that developed and commercialized the
concepts of CADD. In the 1980s, Autodesk (maker of
AutoCAD) and Bentley Systems (Microstation) were
founded and led to the wider availability of CADD on
personal computers. Later in that decade, Parametric
Technology Corp. produced a 3-dimensional design
system.
6.1.3 CIS
GIS represents computerized systems for the storage,
retrieval, manipulation, analysis, and display of
geographically referenced data (Mark, 1997a).
Though the term GIS was first coined by Roger
Tomlinson, director of the Canada GIS in the early
1960s, many of the concepts of GIS lie in the earlier
fields of mapping and cartography. There are several
fields and institutions that contributed to the GIS area
in a non-linear manner over the past 40 years,
resulting in the very powerful and widespread use of
GIS today. Figure 6-3 shows the typical inputs and
results of current GIS packages.
The development of the Geographic Base File/Dual
Independent Map Encoding (GBF-DIME) files by
the U.S. Census Bureau in the 1960s was the first
large-scale use of digital mapping by the govern-
ment. This system led to the production of the
Census Topologically Integrated Geographic
Encoding and Referencing (TIGER) files. Important
geographic work was also being done at universi-
ties throughout the 1950s and 1960s. A grid-based
mapping program called Synagraphic Mapping
(SYMAP), developed at the Laboratory for Com-
puter Graphics and Spatial Analysis at the Harvard
Graduate School of Design in 1966, was widely
distributed and served as a model for later systems.
Output from SYMAP was on a line printer. A
companion program called SYMVU allowed for
mapping of topographic and other data using a pen
plotter.
6-4
-------
A Reference Guide for Utilities
Streams
Point Names
Town
Populations
No.Teenagers
Per Square Mile
Location Map
No. Of People Over 62
Within 2 Miles Of Stream
Figure 6-3. Typical Inputs and Results of Current GIS Packages.
From an application viewpoint, much of GIS model-
ing technology, as it is used today, is largely an
outgrowth of planning approaches that are based on
the work of Ian McHarg, as articulated in his book
Design with Nature (McHarg, 1969). His manual
methods involved overlaying a grid on the area to be
studied, comparing and combining values for
different types of attributes in a grid cell to determine
the suitability of each grid cell for various uses.
Attributes could include characteristics such as land
slope and soil attributes.
He demonstrated this process in his book by creating
maps of different attributes on transparencies with the
darkness proportional to the degree to which that
attribute would support a particular use. For example,
significantly sloping land would be represented as
dark areas because it is difficult to build under these
circumstances. Then the reader could physically
overlay the transparencies and select the lighter areas
which were most appropriate for development. When
computerized, this became the basis of the common
overlay analysis of GIS technology, which served as
the basic modeling technology of GIS for many years.
Another source for GIS technology was computerized
photogrammetry and surveying. In the early 1960s,
the desire to manipulate spatial data in a computer led
to such well-known civil engineering programs as
Coordinated Geometry (COGO), developed at MIT for
calculating surveying analyses on coordinate data. At
the same time, the concept of the Digital Terrain
Model (DTM) was developed, in which the computer
would be used to store a digital database representing
the earth's surface (see the next subsection for more
details on DTM).
In the past decade, development within the GIS
community has been primarily associated with
commercial enterprises that develop and market GIS
software. Companies such as ESRI, Smallworld,
Intergraph, Bentley, Maplnfo, and AutoCAD domi-
nate the GIS field today. Early important public
domain or academic GIS packages such as Geographic
Resources Analysis Support System (GRASS),
developed by the U.S. Army Corps of Engineers and
Clark University, have been largely replaced by the
commercial software packages.
6.1.4 OEMs
OEMs and DTMs refer to representation of ground
surfaces in a computer. Over the past 40 years, they
6-5
-------
A Reference Guide for Utilities
have been used to support applications such as
highway design, sewer design, hydrologic analysis,
mining calculations, and military applications (such
as part of a guidance system for missiles). Various
methods utilizing regular/irregular grids and triangu-
lated irregular networks (TIN) (Mark, 1997b) have
been employed to provide efficient representations of
surfaces. DEMs have become a regular feature of
today's GIS packages. DEM databases are readily
available for most of the U.S. from USGS and other
sources. Figure 6-4 depicts a DTM output.
»
Figure 6-4. Digital Terrain Model of Mount St. Helens after
Eruption in 1980 (R. Home, 2004).
6.1.5 Database Management Systems
Spatial data is composed of two forms of information:
geographic coordinates and attribute information. As
an example, in order to represent a water main,
geographic information describes the coordinates of
the start and end of the pipe and any curves or bends
in the pipe. Attribute information may include the
pipe diameter, length, material, age, and other data of
interest. Typically, attribute data are stored in a
relational database system that may be part of a GIS or
an asset management system. The Relational
Database model for database design was invented by
Dr. E.F. Cobb in 1969 and published in Computer
World in 1985. A relational database system is
composed of a series of tables that are related through
keywords. This model is considered to be highly
efficient and minimizes errors.
6.1.6 Facility Management
Facilities management (or asset management as it is
frequently called today) pertains to use of computer
database and mapping technology to store and
manage information related to physical assets in a
water system. In the 1980s, the term AM/FM was used
to describe the automation of mapping and the
management of facilities represented on those maps.
This typically involved the integration of CAD
technology and database management technology. In
addition to the water industry, AM/FM was used by
the electric and gas industry, telecommunications
industry, and other industries that maintained
physical networks. In the 1990s, the focus of facili-
ties management both shifted and expanded to
encompass a broader management of geographic
spatially arrayed data that are used and maintained by
the various types of utilities. This shift brought a
closer interdependence to the GIS field and frequently
this broader area is now referred to as AM/FM/GIS
(Cesario, 1995) or even more broadly as Geospatial
Information Technology.
6.2 GIS Principles
Understanding the basic principles behind geographic
information systems is difficult because of the breadth
of the field, the rapid change in technology, and the
lack of standardization for terminology. This section
provides a general overview of the most significant
GIS principles.
6.2.1 GIS Features
A GIS is composed of a group of objects or features
that have both a locational description and a descrip-
tion of their characteristics or attributes. For example,
a water tank can be identified by its location in terms
of latitude-longitude or other coordinate systems and
its characteristics, such as diameter, height, and type
of construction. Similarly, a pipe can be described by
its route, diameter, length, material, and age. More
importantly, these attributes can be stored, updated,
and analyzed in a database over time.
Geographic features are stored in three general ways:
vector, raster, or TIN. Though a geographic feature
can frequently be stored in more than one way (e.g., as
a vector or as raster), typically there is a preferred way
to store each piece of information. Under the general
area of vector representation, features can be stored as
points, lines, or polygons. Figure 6-5 illustrates the
three types of features in an example map of a water
POLYGON
LINE
POINT
•-
Figure 6-5. Map of Pressure Zone Showing Three
Types of GIS Vector Data.
6-6
-------
A Reference Guide for Utilities
distribution system pressure zone. Points are geo-
graphically represented as a single set of coordi-
nates in two or three dimensions. A line can represent
either a single, straight-line segment, identified by the
coordinates of the two end points, or a series of
connected line segments to represent a curved line. A
polygon is defined by a closed set of line segments
and identifies the area contained within the defined
outer boundary.
Raster data most commonly refers to a set of data that
has been defined in terms of a regular square or
rectangular grid system that is tied to a geographical
coordinate system. Each grid cell can have one or
more characteristics assigned to it. As an example of
a raster data set, Figure 6-6 illustrates land use
information derived from a satellite that is represented
as a raster database. Other methodologies for storing
raster data include scan lines and other regular grid
cell configurations.
Figure 6-6. Regional Land Cover Characterization
as a Raster Database (USGS, 1992).
The third general type of GIS feature is the TIN
structure. As the name implies, a TIN structure is
composed of a series of irregularly-sized triangular
cells. TIN is most frequently used as a mechanism for
storing topographical information, though it can also
be used to store other discrete or continuous spatial
data fields. The applicability of the TIN structure for
storing topographical information lies in the simple
geometric axiom that three points define a plane, as
illustrated in Figure 6-7. As shown, a continuous
surface can be represented by a faceted set of triangles
with the sides of triangles representing topographical
elements such as streams, ridges, and drainage divides
(Grayman et al., 1975).
The resolution and accuracy of a TIN database
generally depends upon the size of the triangles
relative to the degree of detail in the surface being
represented. Various mathematical techniques can be
used to construct a TIN representation with the most
commonly used method for constructing a TIN from a
series of points known as Delaunay triangulation
(named after a Russian mathematician who invented
the procedure in 1934).
Within a GIS, features are organized as separate layers
in a manner analogous to the original concept
developed by Ian McHarg over 35 years ago. When
viewing GIS data, layers can be turned on or off or
moved forward or backward in order to better under-
stand or view the spatial relationships.
6.2.2 Topology
An important characteristic of GIS is the concept of
topology. Topology may be described as the
locational interrelationship between features. Terms
such as adjacency, intersection, and connectivity are
all topological characteristics that describe how
individual features interact. When we look at a map,
our eyes and mind construct the topological linkages
between features. Observations such as the
Mississippi River forming the boundary between
Illinois and Missouri, the Monongahela River and
Allegheny River intersecting to form the Ohio River,
or that a highway intersection between Interstate 95
and Interstate 10 is completely contained within the
state of Florida are all statements of topology. GIS
Figure 6-7. Triangulation of Elevation (I) Data.
6-7
-------
A Reference Guide for Utilities
constructs the topological relationship between
individual features, and this capability is used in
various analyses and modeling tasks.
6.2.3 Map Projections, Datum, and Coordinate
Systems
Map projections, datum, and coordinate systems
provide the mechanism for establishing a unique
geographic location for a point on the earth's surface.
As scientists have known for centuries (or even
millennia), the earth is approximately a spheroid, or
in reality, an ellipsoid. However, when we are viewing
maps or utilizing spatial data in a GIS, the earth is
represented as a planar (flat) system. For maps or
plans covering smaller areas, the distortion intro-
duced by ignoring the curvature in the earth's surface
is generally insignificant. However, for maps and
plans covering larger areas, this distortion would be
unacceptable. With the wide availability of regional
or national GIS databases, the necessity for accurately
determining coordinates is paramount.
The mechanism for converting a location on the earth
surface to a flat surface is performed using a map
projection. A map projection is a mathematical
relationship for performing this conversion. There are
many projections or relationships that can be used to
make this conversion. Some of the more commonly
used projections include:
• State Plane Coordinates (SPC),
• Universal Transverse Mercator (UTM),
• Albers Equal Area,
• Lambert Conformal Conic, and
• Space Oblique Mercator.
In each projection, the earth is divided into a series of
zones. A best-fit, separate, planar coordinate system
is established for each zone. When examining an area
that straddles multiple zones, such as use of the SPC
system with a metropolitan area that is in multiple
states, coordinate conversions are needed in order to
view the entire area in a single, consistent coordinate
system.
A final important issue in understanding projections
and coordinate systems is the concept of a datum.
Because of the complexity of the shape of the earth
and the inability to exactly describe it mathemati-
cally, the earth has been historically modeled by a
best-fit ellipsoid. The parameters of that ellipsoid are
defined by key datum points located on or in the
earth. For many years, the North American Datum,
developed in 1927 (NAD27) that uses a point on the
earth's surface in Meade's Ranch, Kansas, as an
anchor, was the major standard. Most projections for
North America used this datum.
With improved mathematics and measurements of the
earth, the North American Datum of 1983 (NAD83)
was developed with a datum located within the earth.
This has become the new standard for projections. As
a result, a point having a particular coordinate using
the NAD27 datum may be shifted by tens or hundreds
of feet from a point with the same coordinate using
the NAD83 datum. If this is not properly accounted
for in a GIS system, a map that used the NAD27 datum
would not properly overlay on a map using the
NAD83 datum. There are several GIS utilities
available that will properly convert datasets from one
projection and datum into another, as well as some
newer GIS programs that re-project datasets with
different coordinate systems "on-the-fly." In either
case, it is the responsibility of the GIS user to know
the projection and datum associated with each data
source and to make the appropriate definition.
6.2.4 GIS Database Design
GIS concepts and software provide an opportunity
and a platform for utilizing spatial data. However, in
order to effectively store, analyze, and display the
data, they must be arranged in an organized manner.
Factors that affect database design include: the goals
of the GIS implementation, the short- and long-term
plans for the GIS, the type and number of users for the
particular application, any existing industry-wide
standards, and other application-specific factors.
Zeiler (1999) discusses the issues associated with
database design. Various industry groups are attempt-
ing to define generic data structures for a particular
industry (such as the water utility industry) in order to
facilitate data transfer and common usage of a GIS in
that industry (Grise et al, 2000).
6.2.5 Management of GIS
In the early days of GIS development, the GIS was
typically developed, managed, and used by a central
core of a few people at a governmental or private
organization. With the growth and acceptance of GIS,
there are now frequently many GIS stations using a
specific GIS at an agency, utility or consortium of
utilities. Management of such a system and controls
on the manner in which changes in the GIS are made
are very important issues.
GIS installations may be classified as a personal (or
local) system or an enterprise system. In a personal
system, the GIS is used and managed by an individual
or a small group. On the other hand, an enterprise
system may be used by dozens of users distributed
throughout an agency and many locations. Though
many of the management issues may be similar in
these two scales of operation, the enterprise system
presents a more challenging situation in terms of
6-8
-------
A Reference Guide for Utilities
managing the system. Issues such as personnel (user)
assignment for changing data or backing up the
system and interconnectivity between stations and
users must be carefully spelled out in order to insure
the integrity of the system.
A common management model for large systems
involving multiple users and locations is the combi-
nation of a central enterprise GIS with multiple local
GIS installations. A central core of managers, who
control any modifications to the database and
maintain its integrity, maintains the central system.
Local stations can either access the enterprise system
on a read-only basis or can download copies of all or
part of the database for their local use and modifica-
tion. Specific protocols are then used if changes
made at the local level are to be incorporated into the
enterprise system. These protocols may include
assigned responsibility at the local level to selected
layers within the enterprise system. This form of
management is relatively common with a county-wide
or multi-utility agency managing an enterprise GIS
and a water utility maintaining responsibility for the
water system layers within the GIS.
6.3 Geospatial Data Management
in the Water Industry
Because of the spatial extent and nature of water
supply systems, management of geospatial data is an
important task. This is accomplished through a series
of systems under the overall umbrella of SDMS used
to collect, store, and employ these spatial data. In
some cases, these various systems are integrated; in
other cases, they are independent systems.
6.3.1 CADD
CADD systems have long served as the basis for
designing water distribution systems and facilities
and for managing maps of the water system. Most
utilities and consulting engineers use commercial
packages such as AutoCAD, Intergraph or
MicroStation. The CADD system may be organized
around a collection of maps or plans with a local
coordinate system for each plan or may utilize a
regional coordinate system such as SPC. Many water
utilities use water distribution system models that are
integrated with CADD packages.
6.3.2 GIS
GIS has made significant inroads in supplementing or
replacing CADD packages at many water utilities. GIS
capabilities to store, access and map data are leading
to increased usage of GIS in areas such as planning,
facilities management, and management of customer
and water quality data. Some water utilities share a
GIS database with other entities, such as city or
county governments, and other utilities, such as gas,
electric, and telephone. At many utilities, GIS
technology has also subsumed the capabilities that
were formerly classified as AM/FM systems. Simi-
larly, GIS systems may include an LIS as a means of
storing land property, parcel, and ownership informa-
tion and geographic descriptions. OEMs are also a
regular feature in GIS packages. They provide a
mechanism for storing topographical information. In
the past few years, integration of GIS with water
distribution system models has been a significant area
of research and development in the water industry.
6.3.3 CIS
CIS provides a mechanism for storing and using
information on water consumption by customers. The
geographic component in a CIS is an address and/or a
geographic coordinate. AMR systems facilitate
collection of consumption data that can be stored in
databases. Standard GIS "address matching" capabili-
ties facilitate conversion of addresses to geographic
coordinates. A geographically enabled CIS provides
an excellent mechanism for automatically recording
current consumption data to be used in water distribu-
tion system models.
6.3.4 SCADA
SCADA systems typically include capabilities to
remotely access information on the state of the water
system, to manually or automatically control compo-
nents such as pumps and valves, and to store and
display current or historical time-series data about
system operation. A wide range of commercial
SCADA hardware/software systems is available and
can be tailored to the specific needs of the water
utility. Each component that is referenced in a SCADA
system can have a unique geographic identifier that
can be used as a linkage to a GIS or other spatial data
management systems. Research and development is
underway related to integrating SCADA systems and
hydraulic/water quality distribution system models so
that these models can be used in real-time operation
and emergencies.
6.3.5 LIMS
LIMS are computerized systems for managing samples
in a laboratory. Such systems typically include a
mechanism for storing, managing, displaying, and
tracking samples. Since the origin of a sample must
be identified both spatially and temporally, this
information provides a means of associating LIMS
data with other spatial database management systems.
6.3.6 Support Technology
Other technological advances related to spatial
database management that are used by water utilities
include GPS and RDBMS. GPS is a widely used
technology in surveying and can be used for tagging
field data with a geographic coordinate. RDBMS is a
6-9
-------
A Reference Guide for Utilities
general methodology for efficiently storing informa-
tion as a series of related 2-dimensional tables. Most
modern database management systems associated
with CIS, LIMS, and other systems utilize the RDBMS
structure.
6.4 Integration of Geospatial Data
Management and Modeling
The concept of integrating water distribution
system modeling with geospatial database manage-
ment systems has been evolving over the past
quarter of a century and continues to be a major
focus of development in the water industry today.
Early water distribution system models were stand-
alone entities. In the very early models, input was
provided by punch cards and output was in the
form of printed tabular information. This cumber-
some I/O gave way to input via terminals in the
1980s and GUIs in the 1990s. The 1990s also saw
the first commercial integration of water distribu-
tion system models with CADD followed by
integration with GIS in the 2000s.
The basis for integrating water distribution system
models with geospatial data can be traced back to an
early study that interfaced a planning level sewer
design model with a TIN-based GIS called ADAPT
(Areal Design And Planning Tool) (Grayman et. al.,
1975). This approach was called "geo-based model-
ing" and was subsequently applied to various other
water engineering situations such as hydrologic
modeling (Males and Gates, 1979). In these systems,
a geo-based network representing sewer lines or
streams was integrated with GIS elevation, land use,
and soil data. This network directly interfaced with
design and simulation models.
In the 1980s, the same geo-based modeling concept
was applied to water distribution system analysis
through a series of EPA research projects. The Water
Supply Simulation Model (WSSM) integrated a geo-
based, link-node system to several models including a
hydraulic model, a steady-state water quality and cost
allocation model, and various display and editing
routines (Clark and Males, 1985). Subsequently,
WSSM was expanded to include an interface to GIS
files using AutoCAD. USGS digital line graph (DLG)
files of road networks and OEMs were used within
AutoCAD to create a detailed representation of the
water distribution system. The resulting database was
used to generate an input file for the Wadiso hydrau-
lic model whose engine worked as a prototype for
EPANET
In the past 10 years, commercial vendors of network
modeling software working in conjunction with
CADD and GIS vendors have led the integration of
modeling software and spatial database technology.
In the mid 1990s, hydraulic/water quality models
were built to operate within AutoCAD. More recently,
commercial modeling systems have been released as a
version that are fully integrated and operate within
the GIS environment. Commercial products include
WaterGEMS (Haestad Methods/Bentley Systems) and
InfoWater(MWHSoft).
6.4.1 Model Integration Taxonomy
The term "integration" can refer to a wide range of
capabilities related to use of network models in
conjunction with a spatial database system. Shamsi
(2001) provides a taxonomy of three levels for model-
database integration. These are described as inter-
change, interface, and integration.
Interchange provides a mechanism for transferring
data between a spatial database such as GIS and a
model. With interchange, there is no direct linkage
between the two systems. Rather, they are run
separately and information is extracted from one
system and stored in an intermediate file that is
subsequently accessed by the other system. In the
direction of database to model, information stored in
a GIS is used to generate a complete or partial dataset
that is used as input to the model. In the other
direction, output from a model is used as input into a
GIS in order to display the results of the model
application. Most commercial water distribution
modeling packages can interchange data with CADD
and GIS platforms.
An interface involves a direct connection between the
database and the model in order to transfer informa-
tion in either direction. As is the case in interchange,
the two systems still operate independently, but in
this case, there is a direct linkage so that intermediate
files are not necessary. Protocols and structures must
be established and compatible within the two systems
in order to support this interface. Current trends are
directed towards open architecture in which informa-
tion on the data structures for models and GIS are
designed to data structures. For example, H20Map
and WaterCAD are standalone software packages
which can directly interface with data in CADD and
GIS platforms.
True integration is the most sophisticated of the three
methods. Ideally, the two systems work together
seamlessly as a single entity. In such integration,
either the model can operate within the spatial
database software or the spatial database capability
can be part of the model. For example, WaterCAD and
H20NET software packages are integrated and
operate within AutoCAD. Info Water and WaterGEMS
software packages are integrated to operate within
ArcGIS.
6-10
-------
A Reference Guide for Utilities
6.4.2 Issues in Integrating GIS and Water
Distribution System Models
As an evolving technology, there are still issues in
truly integrating GIS technology with water distribu-
tion system models. These issues primarily revolve
around the level of detail required in the two systems
and the procedures for updating the model and the
database.
In most cases, a water utility GIS is used for many
purposes including mapping, facility management,
planning, and modeling support. As a result, there
may be a great deal of detail in the GIS. For example,
it may include hydrants, shutoff valves, water meters
and household connections, air release valves, and
other appurtenances. On the other hand, typically,
water distribution system models do not explicitly
include many of these components. This is shown
graphically in Figure 6-8. In this case, only junction
nodes and pipes are included in the model representa-
tion. As a result, the GIS representation includes 17
links and 11 nodes, and the model representation
includes 6 links and 3 nodes. The disparity between
the two representations can increase by another order
of magnitude if water meters and customer connec-
tions are included in the GIS. Various approaches are
taken to deal with this situation.
• A very detailed model is constructed that
includes all of the elements in the GIS. This
solution can result in a very large model with an
excessive number of nodes and links.
• The GIS representation goes through a
consolidation (skeletonization) procedure to
eliminate unneeded nodes and to aggregate the
resulting links in order to construct the model
representation. Though this results in a more
appropriate model, it adds an intermediate step
between the GIS and the model. Additionally,
after the consolidation process, there is no
longer a one-to-one correspondence between
GIS and model features. This lack of
correspondence leads to issues related to storing
model output in the GIS and updating the model.
• Multiple representations are maintained within
the GIS for different uses. The detailed
representation is the complete, base case and
used for facility management while the
skeletonized version is used for modeling. This
approach has the limitation that requires
changes in information to be made in multiple
databases.
• The basic link-node network (as used in the
model) is maintained as the base case in the GIS
and associates other components (such as
hydrants) with links rather than structurally
embedding them in the network.
17 links - 11 nodes
Figure 6-8a. Typical Representation of a Pipe Section in GIS.
6 links - 3 nodes
Figure 6-8b. Typical Representation of a Pipe Section in a
Network Model.
In any of the options described above, procedures for
updating the GIS are essential. There are many issues
associated with GIS updating such as authorization of
specific users to make changes, nature of the changes
as permanent or part of a 'what-if modeling scenario,
and frequency of replication of the two databases if
separate model and GIS files are maintained. All of
these should be carefully spelled out prior to design-
ing and implementing a GIS.
6.5 Use of GIS in Water Utilities -
Case Studies
This section presents the potential uses of GIS in the
water utility industry. Case studies from the Las Vegas
Valley Water District (LWWD) and Denver Water are
presented.
6.5.1 Use of GIS at LWWD
Between 1989 and 2004, Las Vegas grew faster than
any other metropolitan area in the U.S. As a result,
LVVWD has more than doubled its service area
population during this period. In 1989, the service
area population was 558,000 and in 2004 it rose to
1,209,000, representing an increase of 651,000 people
serviced by LWWD (Jacobsen and Kamojjala, 2005).
Figure 6-9 is a GIS representation of the LWWD
distribution system growth between 1989 and 2004.
To address a variety of issues related to this rapid
growth, LVVWD integrated the functions of master
planning, operational planning, and development
review by integrating its GIS data with modeling,
SCADA, and enterprise data (such as CIS, AM/FM and
LIMS). Figure 6-10 presents the conceptual relation-
ship model of these functions and potential integra-
tion benefits (Jacobsen et al., 2005).
During the process of integration, LVVWD developed
a one-to-one relationship between the GIS spatial data
and its network model (Jacobsen and Kamojjala,
2005). An example of this one-to-one relationship is
6-11
-------
A Reference Guide for Utilities
Master
Planning
increase
Operational
Efficiency
oost-Effective
and Timely
New Facilities
Operational
Planning
Model SCADA
Spatial
Maintain/lmpro
Level of Service
to Customers
Development
Review
Figure 6-9. LWWD Distribution System Growth.
Figure 6-10. Conceptual Relationship Model for
Integration.
shown in Figure 6-11. Depending on the size of the
network, developing such a relationship and subse-
quent data integration has both advantages and
disadvantages.
^ Information
nstallation Year
RetirementYear
Zone
Material
Lining
Cost ID
B Fife Create Edl View Exchange Scenano lools Utities Window Help
D I* H |ALLPIPE2300, Allpipe 2300 Zone
Figure 6-11. One-to-One Relationship Between GIS and Network Modeling Data.
6-12
-------
A Reference Guide for Utilities
The advantages include ease of search and retrieval
with other data/applications, and ease of importation,
development, and maintenance. For a large network
model, the disadvantages include an increase in the
runtime of the network model due to the addition of
detailed components and relatively slow water quality
simulations. To minimize this, LWWD has taken an
"all-pipes capable" approach where the distribution
system is divided according to existing pressure zones
and attached to an operational backbone network
(skeletonized). Each of the zone models can be
attached seamlessly to the backbone network for
detailed hydraulic and water quality modeling.
Examples of how GIS data are used by LWWD on a
day-to-day basis are presented below.
6.5.1.1 Pressure Complaint Resolution
Once a pressure complaint is received from a cus-
tomer, the GIS data is searched for parcel and account
information, together with modeled and measured
pressure data in the vicinity of the complaint. Figure
6-12 shows an example search window. Depending on
the results of the analyses, a crew may be dispatched
to trace field pressures and abnormal conditions from
the water supply source to the customer location, and
to install hydrant pressure recorders to capture
dynamic variations. Upon retrieval of field informa-
tion, model and field results are compared to identify
possible problems. Figures 6-13 and 6-14 present
examples resulting from this search (Jacobsen and
Kamojjala, 2005).
6.5.1.2 Water Main Break Analysis
During a water main break, it is critical to quickly
identify the distribution system valves that must be
closed in order to minimize water loss, potential
flooding, or possible contamination. Repairs must be
performed quickly so that service can be resumed.
Figure 6-15 illustrates the procedure for rapidly
identifying the valves to be isolated utilizing the GIS
and modeling tools. A list of affected customers is
generated for appropriate notification (Figure 6-16).
An analysis is performed to evaluate the impact on
existing services. Figure 6-17 shows a comparison of
pressures after a shutdown and identifies a lower-
pressure area after the shutdown. Response to
emergencies, such as main breaks, can be provided
quickly and accurately using integrated GIS tools
(Jacobsen and Kamojjala, 2005).
6.5.2 Geo-coding for Demand Forecasting and
Allocation at Denver Water
Between January 1997 and December 2000, Denver
Water conducted a treated water study to evaluate the
transmission, pumping and storage system for
capacity and the need for new facilities for capital
planning and operations. Denver Water made
extensive use of GIS tools for demand forecasting and
» £* ign Bart SeMwi I«* HMo« U*
* * BASE. Be*B Data Scetieva
— Ottrfcvwn
— D&rtiuaon (OZ)
- DiartouBon • Bypass
— fn Hydrant Lateral
— Fre Hydrant Lateral (OZ)
— Other/unknown
— Overfbw/oram
— Pumpng sta&xi Inw/Outjet
— Reservoir Uter./ outlet
— Reiervor biW / Outlet (OZ)
— Sample Ln*
— servrce Later*
— stub
— Surge Supresean
— Transm&son
(OZ)
— wash
— WKte/r>«tisrge
-welCMeetor
-zone wet /outfet
D Account
O
a AHFNSDe.Hydrant
• Red: BM4J
• Green: Band_2
• B*je: Band_3
D ptequad pdyoan
CorMow I Op«a»or< Annotation
output data.. Dona.
Drawftg » H a - A - :"-: |£Arial
Figure 6-12, Pressure Complaint Resolution - GIS Parcel/Account Search Window,
6-13
-------
A Reference Guide for Utilities
P€AlhCTEA
PARCEL.
PARCELJO
OLDFO.
PARCEL
POLOA
PARCELJJPD
C£NTB«D_CHG
AREA CH'L
CAIC.ACRES
ASSR_ACRES
PO^PROBLEM
FOSUSD
POWO
PONOX
OBJECTID
SHAPE
SHAPE AREA
SHAPE.LEN
— Transmsson (02)
— Transmission - Bypass
— wash
— Wane (Cfcchaige
— weicoteaor
— Zone W«/Outlet
- a Account
O
Red: Band.l
Green:
Sue: Sand_3
D pbqudd pctygon
Figure 6-13. Pressure Complaint Resolution - Parcel and Hydrant Location.
BT]
' Pl'lri- I - luljiull -,1^
Field Value
SDE Parcels APEA
M I : - .
PARCEL
POQA 35
PARCEUJPD S/3/1993
CENTROC_CHG 0
AREA.CHG -1
C*LC ACRES D15477
ASSRJVCRES D
PO,PROBLEM 1
POSUBD 11
PCLNO DBS
PCLNOX OK
OBJECT1D
SHAPE
SHAPEAREA 6742M3105
— Transmsson (O2)
— Tfansmsson - Bypass
— Wash
— Waste/
— Zone W«/ Outlet
D Acraunt
O
• Red:
• Green: Bandj
• Bine: Bdr<3
P [:S--iitai1
Figure 6-14. Pressure Complaint Resolution - Model and Field Pressure Comparison.
6-14
-------
A Reference Guide for Utilities
9X1
ltd
*\^
Call out
ta-ir
WNBC-
6'-0"
EMIBC
Quad
Valve
No.
i :-,:-14-
35
District
Mop
12&-34-
SE
Address
6UB1 N
Rainbow
Dlvd
Dlamet>r
16
Drawing
XL-44W-
006
Normal
Portion
u
Currant
Position
j
Status
busting/
Active
Valve
Type
butterty -
West
Jurlsdlct
ion
oounty -
Clark
OA Map
Purvey
Grade
GPS
Data
Install
date
4/rjm
Critical
Service
t
No
Vault
Mo
Vault
Manufac
turer
Mueller
Emergency Work Order
Task:
Date:
Valve No.
Address:
Drawing:
ID Feature:
Size:
Vault:
Close in-line valve
11/11/1234
125344-35
50S1 N Rainbow Blvd.
XC-4462-006 (attached)
353622
16-inch
No vault
Comments:
Urgent task and require completion by 5:30PMon 11/11/1234.
! 1 ! ! I 1 1 1 1
Drawing: XC-4462-006
Figure 6-15. Water Main Break Analysis - Valve Isolation.
fte &H Sfiew insert Detection l«fc jMndcw UHp
Pumpmg Statwn Inltt/OutkS
ReparOevKe
Reserve* mtet/outi«
Reset vor mitt/ Outlet (OZ)
Sample Lne
- servte Lmerai
Stub
— Transmisson (OZ)
— Tignsmsson - Bypass
— wash
— Waste /Discharge
— Zone Wet / Outlet
ft»FMSO€-H«er
a WORESSOE.PreKure_Zooes
a
3 KSDBAJ1EHALPHOTC
• 16' WEEB
_|24' WEEB
6' EEEB
4' WEPL
27'EWPL
7- 8" £ OF FH
15' EWEB
23' 6' WEEB
$ EEE6
7" iVEEB
17'CWE6
ff EEEB
21'WEEB
2S1 8- EWPL
5' 6" EWEB
1 7' 4- WEES
11' 8" WEEB
23' EWEB
is1 ewes
2 6- WEEB
J2VWEEB
<
00012
00012
00012
00012
00012
00012
00012
00012
N
N
H
N
N
N
N
1
J317 EL PARGUE AVE
4313ELPARQUEAVE
Wl 7 W OAKEY BLVD
«1 3 W OAKEY BLVD
4309 W OAKEY BLVD
4305 W OAKEY BL VO
4301 W OAKEY BLVD
4217 W OAKEY BLVD
dj13 W OAKEY B^VD
| Affected Accounts 1
00012
000)2
00012
00012
00012
00012
'00012
00012
00012
N
N
N
N
N
N
5/8
W
sm
51B
&8
109 w OAKEY 5LVD
1 05 W OAKEY BLVD
101 W OAKEY BLVD
100 LAS LOMAS AVE
10J LAS LOMAS AVE
1 0B LAS LOMAS AVE
112 LAS LOMAS AVE
116 LAS LOMAS AVE
4200 LAS LOMAS AVE
i
I
O.Wft
OMiUWi
Oased
4200 in
I30.0B
jpoT] • / n A,'
-•----•--•
- •- - -
Figure 6-16. Water Main Break Analysis - Impacted Customer List.
6-15
-------
A Reference Guide for Utilities
S * v £>«>
* - *•!] B • si
fl
- 3 Junction
PRESSURE
-29.090000 - 0.000000
0.000001-19.!
O 20.000000 • 39.999999
40X00000 - 90.000000
30.000001 - lOO.OOOQQfl
a Tank
a
S Reservoir
D jlMFMSW.WatotJTO
— oloflw values*
Subtype
— Artftfal Recharge
— AV/AA Lateral
— AV/AR lateral (OZ)
— Bbw Off Lateral
PRESSURE
-29.090000 - 0.000000
O 0-000001 -19.999999
O 20.000000 - 39.999999
O 40.000000 - 80.000000
-X":M;:- 100.000000
- Dttrtajtlon (OZ)
— DttnbutOn - Bypass
— f*e Hydrant Lateral
— Fre Hydrant Lateral (OZ)
-Other/Unknown
— Overflovrforain
— Pumping Station Inlet/Outlet
Data
— Reserves mtrt / Oubet
-Resefvorlnfet/Outlet
— Sample Lne
-Service
;r ir: «r g»& |[p^ rf dt |
0-29.090000-0.000000
0.000001-19,999999
O 20.000000 • 39.999999
40X00000 - SO-000000
$0,000001 -100.000000
Tank
—
-------
A Reference Guide for Utilities
allocation as part of this analysis. Specifically,
existing demands were spatially allocated using
address geo-coding (Strasser et al., 2000).
Denver Water has been a completely metered
system since 1992. Historic consumption data is
available from 1993 to the present. Consumption
data was extracted from the billing system, and
imported into an MS Access database for use in
various tasks including the treated water study.
The extracted data included both customer address
and customer class information (e.g., single family,
multi-family, commercial, industrial, and public).
The customer class information allowed the
demand information to be aggregated by customer
class. Because the service area was large (over 250
square miles), it was important to identify the
location of the consumption demand points. This
is where the use of GIS became very important
(Strasser etal., 2000).
Denver Water used the "address geo-coding" feature
available within Arclnfo which allowed for each
customer or demand point to be identified on a base
map. Using this process, a dot is placed on the base
map representing each customer that could be
positively geo-coded. The match rate was over 93
percent. Those accounts that could not be geo-coded,
mostly large accounts representing master meter
accounts, and wholesale customers, were entered in
the system manually. Results from the geo-coding
process for one pressure zone are illustrated in Figure
6-18. A quality control check was performed on the
results of this geo-coding process by reconciling
consumer demands with Denver Water's annual
statistical report (Strasser et al., 2000).
6.6 Summary
Use and management of geospatial data is an impor-
tant aspect of the design and operation of water
systems. This can be accomplished through a range of
systems under the overall umbrella of SDMS utilized
to collect, store, and use the spatial data. This
umbrella covers not only the broad topics of GIS and
CADD that are widely recognized as geospatial data
systems, but also systems such as SCADA and LIMS
that have a spatial component associated with all
data.
The area of spatial database management is continu-
ing to evolve within the water industry. Just as the
capabilities of the various individual components
within the SDMS umbrella continue to expand, the
integration of the various systems is an active area of
development. Water distribution system analysis is a
significant beneficiary of these improvements and
integration. As a result, models can be built more
quickly and in greater detail. Information on facili-
ties and demands can be routinely updated. The
results of a model application can be rapidly dis-
played and viewed along with other spatial data. The
prospect of real-time application of models to assist in
system operation under routine conditions or under
emergency conditions is getting closer.
Figure 6-18. GIS Geo-coding - Metered Sales Demand
Allocation Procedure.
6-17
-------
A Reference Guide for Utilities
References
Clark, R.M., and R.M. Males. "Simulating Cost and
Quality in Water Distribution." Journal Water
Resources Planning and Management, ASCE.
lll(4):454-466. 1985.
Cesario, A.L. Modeling, Analysis, and Design of
Water Distribution Systems. AWWA, Denver, CO.
1995.
Grayman, W.M., R.M. Males, WE. Gates, and A.W.
Hadder. "LandBased Modeling System for Water
Quality Management Studies." Journal of Hydrau-
lics, ASCE, HY5. May 1975.
Grise, S., E. Idolyantes, E. Brinton, B. Booth, and M.
Zeiler. "Water Utilities ArcGIS Data Models." ESRI
Press, Redlands, CA. 2000.
Home, R. "Gallery of 3DEM Terrain Images." http://
www.visualizationsoftware.com/3dem/gallery.html.
2004.
Jacobsen L., and S. Kamojjala. "Full System Models
and GIS Integration." AWWA Annual Conference and
Exposition, San Francisco, June 2005.
Jacobsen L., S. Kamojjala, and M. Fang. "Integrating
Hydraulic and Water Quality Models with Other
Utility Systems: A Case Study." AWWA Information
Management and Technology Conference, Denver,
April 2005.
Males, R.M., and WE. Gates. "ADAPT: A Digital
Terrain Model-Based Geographic Information
System." Symposium, Laboratory for Computer
Graphics & Spatial Analysis, Harvard University,
Cambridge, MA. 1979.
Mark, D.M., N. Chrisman, A. U. Frank, P. H. McHaffie,
and J. Pickles. "The GIS History Project." UCGIS
Summer Assembly, Bar Harbor, ME. http://www.
geog.buffalo.edu/ncgia/gishist/bar_harbor.html.
1997a
Mark, D.M. "The History of Geographic Information
Systems: Invention and Re-Invention of Triangulated
Irregular Networks (TINs)." Proceedings, GIS/LIS '97.
http://www.geog.buffalo.edu/ncgia/gishist/
GISLIS97.html. 1997b.
McHarg, I.L. Design with Nature. Published for the
American Museum of Natural History by the Natural
History Press. 1969.
NASA. "Earth Observatory - Remote Sensing." http://
earthobservatory.nasa.gov/Library/RemoteSensing/
remote.html. Undated.
O'Connor, J.J., and E. F. Robertson. "The History of
Maps." http://geogdata.csun.edu/geogcourses/
history_of_maps.html. Undated.
Queensland Government. "Maps and Compasses -
The History of Surveying." http://education.qld.
gov.au/curriculum/area/maths/compass/html/survey-
ing/suhis.html. Undated.
Radio Shack. "A Guide to the Global Positioning
System (GPS)." http://support.radioshack.com/
support_tutorials/gps/gps_main.htm. 2004.
Shamsi, U.M. "GIS and Modeling Integration." CE
News, 13(6). 2001.
Strasser, A., N. Diallo, and E.J. Koval. "Development
& Calibration of Denver Water's Hydraulic Models for
a Treated Water Study." AWWA National Conference,
Denver, Colorado, 2000.
Sutherland, I.E. "Sketchpad: A Man-machine
Graphical Communications System." Reprinted
version of original 1963 thesis issued as a report by
Cambridge University. UCAM-CL-TR574. Cam- "
bridge, UK. http://www.cl.cam.ac.uk/TechReports/
UCAM-CL-TR-574.pdf. 2003.
USGS. "National Aerial Photography Program
(NAPP)". http://edc.usgs.gov/guides/napp.html.
Undated.
USGS. "Scientific Assessment and Strategy Team
(SAST) Study of the Moberly Quadrangle." http://
edc.usgs.gov/sast/moberly.html. 1993.
USGS. "National Land Cover Characterization
(NLCD) Database Program." http://landcover.usgs.
gov/images/glensfalls_web.jpg. 1992.
Whitfield, P. The Image of the World: 20 Centuries of
World Maps. Pomegranate Artbooks, San Francisco,
CA, 144 p. 1994.
Zeiler, M. Modeling Our World. ESRI Press,
Redlands, CA. 1999.
6-18
-------
A Reference Guide for Utilities
Chapter 7
Real-World Applications -
Planning, Analysis, and Modeling Case Studies
The previous chapters of this reference guide show-
cased several tools for analyzing water quality in
drinking water distribution systems. Some of these
chapters also have relevant case studies that relate to
the individual topic of discussion for that chapter.
This chapter focuses on the broader application of
multiple tools to analyze real-world situations. The
types of applications presented here include: recon-
struction of historical contamination events, analysis
of waterborne outbreaks of infectious diseases,
regulatory compliance, monitoring systems location,
and security. Each real-world application is presented
as an individual case study and the portions related to
water quality and analysis have been highlighted
along with some specifics on techniques used in the
analysis.
7.1 Analysis of Waterborne
Outbreak - Gideon, Missouri
This case study is focused on evaluating the distribu-
tion of microbiologically contaminated water in a
distribution system. The supporting investigations
for this case study were primarily sponsored by the
EPA, CDC, and the State of Missouri.
Key Phrases to Characterize Case Study: water
borne outbreak analysis, salmonella, tank contamina-
tion, hydraulic and water quality modeling, exposure
modeling, contamination assessment, and flushing.
7.1.1 Gideon Case Study Overview
From November 1993 through January 1994, the
Missouri Department of Health (MDOH) had identi-
fied 31 cases of laboratory-confirmed salmonellosis
infections associated with a waterborne outbreak in
Gideon, Missouri (Clark et al., 1996). The State
Public Health Laboratories identified 21 of these
isolates as dulcitol negative Salmonella serovar
Typhimurium. Salmonella is a pathogenic bacterium
that has been classified into several serotypes
(common set of antigens). Salmonella serovar
Typhimurium is among the most common Salmonella
serovars causing salmonellosis in the U. S. Fifteen of
the 31 laboratory culture-confirmed patients were
hospitalized (including two patients hospitalized for
other causes and who developed diarrhea while in the
hospital). These 15 patients were admitted to 10
different hospitals. Seven nursing home residents
exhibiting diarrheal illness died; four of these
patients were culture confirmed (the other three were
not cultured). Two of the patients had positive blood
cultures. Interviews conducted by the MDOH during
this period suggested that there were no food expo-
sures common to a majority of the patients. However,
all of the ill persons, including the culture-confirmed
patients, had consumed municipal water which
supported the association. The MDOH reported their
suspicion to the Missouri Department of Natural
Resources (MDNR).
7.1.2 The Gideon Water System Setup
The Gideon municipal water system was originally
constructed in the mid-1930s and obtained water from
two adjacent, 1,300 ft deep wells. The well waters
were not disinfected at the time of the outbreak. After
the outbreak emergency, chlorination was initiated,
and later a permanent chlorination system was
installed. The distribution system consisted primarily
of small-diameter (2-, 4-, and 6-inch) unlined, steel
and cast iron pipe. Tuberculation and corrosion were
major problems in the distribution pipes. Raw water
temperatures were unusually high for a groundwater
supply system (58°F), because the system overlies a
geologically active fault. Under low flow or static
conditions, the water pressure was close to 50 psi.
However, under high flow or flushing conditions the
pressure dropped dramatically. These sharp pressure
drops were evidence of major problems in the Gideon
distribution system. The municipal system had two
elevated tanks. One tank was a 50,000 gallon (gal)
tank (referred to as small tank) and the other was a
100,000 gal tank (referred to as large tank).
Initially, another 100,000 gal privately owned tank
was suspected to be the cause of the outbreak (as it
was in a state of disrepair) and connected to the city
water system. However, subsequent investigations
revealed that this private tank was connected via a
backflow prevention valve to the city water system
that was later confirmed to be functional. Further-
more, the Salmonella found in a sample collected at a
hydrant matched the serovar of the patient isolate
when analyzed by the CDC laboratory (comparing
deoxyribonucleic acid [DNA] fragments using pulse
field gel electrophoresis). Although the samples from
the private tank sediment also contained Salmonella
serovar Typhimurium dulcitol negative organisms, the
isolate did not provide an exact DNA match with the
other two isolates. No Salmonella isolates were found
elsewhere in the system. Therefore, the subsequent
EPA field investigations and modeling efforts focused
on the two municipal tanks as the source of contami-
nation.
7-1
-------
A Reference Guide for Utilities
7.1.3 EPA Field Study
On January 14, 1994, an EPA field team, in conjunc-
tion with the CDC and the State of Missouri, initiated
a field investigation that included a sanitary survey
and microbiological analyses of samples collected on
site. A system evaluation was also conducted in
which EPANET was used to develop various scenarios
to explain possible contaminant transport in the
Gideon system. Prior to the Gideon outbreak, a
similar waterborne disease outbreak in Cabool,
Missouri, and subsequent advancements in water
quality modeling firmly established the use of water
quality models to analyze such events.
The key analysis was focused on a flushing program
conducted earlier by the utility in response to taste
and odor complaints. A sequential flushing program
was conducted on November 10, 1993, involving all
50 hydrants in the system. The flushing program was
started in the morning and continued through the
entire day. Each hydrant was flushed for 15 minutes
at an approximate rate of 750 gallons per minute
(gpm). It was observed that the pump at one of the
wells was operating at full capacity during the
flushing program (approximately 12 hours), which
would indicate that the municipal tanks were dis-
charging during this period.
During the evaluation, it was hypothesized that the
taste and odor problems may have resulted from a
thermal inversion that had 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 a 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 (Fennel et al., 1974). As a consequence,
the utility initiated the aforementioned city-wide
flushing program. Turbulence in the tank from the
flushing program could have stirred up the tank
sediments that were subsequently transported into the
distribution system. It is likely that the bulk water
and/or the sediments were contaminated with Salmo-
nella serovar Typhimurium. During the EPA field
visit, a large number of pigeons (bird droppings are
known to contain Salmonella) were observed roosting
on the roof of the 100,000 gal municipal tank.
7.1.4 Distribution System Evaluation
The EPA study team evaluated the effects of distribu-
tion system design and operation, demand, and
hydraulic characteristics on the possible propagation
of contaminants in the system. Given the evidence
from the lab samples and the results from the valve
inspection of the private tank, 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
In 1991, a joint workshop sponsored by the EPA and
AwwaRF recommended the application of water quality
modeling techniques to evaluate waterborne disease
outbreaks. The first opportunity to attempt this type of
application arose as a result of an outbreak that occurred
between December 15, 1989, and January 20, 1990, in
Cabool, Missouri, population 2,090 (Geldreich et al.,
1992). During the outbreak, residents and visitors to
Cabool experienced 243 cases of diarrhea (85 bloody)
and six deaths. The illness and deaths were attributed to
the pathogenic agent E. coli. serotype 0157:H7. At the
time of the outbreak, the water source was untreated
groundwater. Shortly after the outbreak was identified,
EPA was invited to send a team to conduct a research
study with the goal of determining the underlying cause
of the outbreak.
Exceptionally cold weather prior to the outbreak
contributed to two major water system line breaks and 43
water meter replacements throughout the city area. The
sewage collection lines in Cabool were located (for the
most part) away from the drinking water distribution
lines but did cross or were near to water lines in several
locations. At the time of the outbreak, stormwater
drained via open ditches along the sides of the streets
and roads. During heavy rainfalls, sewage was observed
to overflow manhole covers, and to overflow streets in
several locations, parking lots and residential founda-
tions.
The Dynamic Water Quality Model (DWQM), developed
by EPA, was applied to examine the movement of water
and contaminants in the system (Grayman et al., 1988).
Steady-state scenarios were examined, and a dynamic
analysis of the movement of water and contaminants
associated with meter replacement and the line breaks
was conducted. Typical demand patterns were developed
from available meter usage for each service connection,
and it was found that the water demand was 65 percent of
the average well production, indicating inaccurate
meters, un-metered uses, and a high water loss in the
system.
The modeling effort revealed the pattern of illness
occurrence was consistent with water movement patterns
in the distribution system assuming two water line
breaks. It was concluded that some disturbance in the
system, possibly the two line breaks or 43 meter replace-
ments, allowed contamination to enter the water system.
Analysis showed the simulated contaminant movement
covered 85 percent of the infected population.
The application of DWQM proved to be a vital step in
completing the analysis of the outbreak. The next
opportunity to apply water modeling techniques oc-
curred in 1994 as a result of a waterborne outbreak in
Gideon, Missouri (Clark et al., 1996). In the intervening
period, EPA had developed EPANET and Gideon
provided an opportunity to test its application.
7-2
-------
A Reference Guide for Utilities
the large municipal tank in conjunction with the
flushing program. Other possible sources of contami-
nation, such as cross connections were also studied.
The system layout, demand information, pump
characteristic curves, tank geometry, flushing pro-
gram, and other information needed for the modeling
effort were obtained from maps and demographic
information and numerous discussions with consult-
ing engineers and city and MDNR officials. EPANET
was used to conduct the contaminant propagation
study (Rossman et al., 1994).
The EPANET network model was calibrated by
simulating flushing at the hydrants assuming a
discharge of 750 gpm for 15 minutes. The "C" factors
(pipe roughness - see Chapter 4) were adjusted until
the head loss in the model matched head losses
observed in the field. After the calibration, the
hydraulic model was simulated for 48 hours. Thereaf-
ter, the flushing program was simulated starting at 8 a.m.
on day 3, by sequentially imposing a 750 gpm
demand on each hydrant for 15 minutes. Utilizing the
TRACE option in EPANET, the percentages of water
from both municipal tanks were calculated at each
node over a period of 72 hours.
During the simulation of the flushing program, the
pump at one of the wells was operated (as previ-
ously observed) at full capacity, which was over
800 gpm, and then reverted to cyclic operation.
The simulation results showed that the tank
elevation fluctuated for both municipal tanks, and
both the tanks discharged during the flushing
program. At the end of the flushing period, nearly
25 percent of the water from the large municipal
tank passed through the small municipal tank
where it was again discharged into the system. The
model predicted dramatic pressure drops during the
flushing program. Based on the information
available, it was felt that these modeling results
replicated the conditions that existed during the
flushing program closely enough to provide a basis
for an analysis of water movement in the system.
Data from the simulation study, the microbiological
surveillance data, and the outbreak data were utilized
to provide insight into the nature of both general
contamination problems in the system and the
outbreak itself. The water movement patterns showed
the majority of the collected samples that were total
coliform and fecal coliform (FC)-positive occurred at
points within the zone of influence of the small and
large tanks. During both the flushing program and for
large parts of normal operation, these areas were
predominately served by tank water, which confirmed
the belief that the tanks are the source of the fecal
contamination since there were positive FC samples
prior to chlorination. Figure 7-1 shows the compari-
son of early confirmed cases of Salmonella positive
sample versus the estimated distribution of tank water
during the first six hours of the flushing program.
Homes called as part of CDC survey
Residences with confirmed case
Hydrant with confirmed Sa/mone//a
Gideon Schools - reflects increase in absentee level
20% or more of Small Tank water
20% or more of Large Tank water
Figure 7-1. Comparison of Early Confirmed Cases of
Salmonella Positive Sample Versus the Estimated
Distribution of Tank Water During the First 6 Hours
of the Flushing Program.
7.1.5 Case Study Summary and Conclusions
Data from the CDC survey of the outbreak, in combi-
nation with the EPANET simulated water movement,
were utilized to establish the possible source of
contamination. An overlay of the CDC data on the
water movement simulations showed that the areas
served by the small and large tanks (during the first
six hours of the flushing period) coincided with the
earliest recorded infectious cases. Furthermore, the
earliest recorded cases and the positive Salmonella
hydrant sample were found in the area that was
primarily served by the large tank, but outside the
small tank's area of influence.
The investigators concluded that during the first six
hours of the flushing period, the water that reached an
infected resident and the Gideon School (the earliest
reported infections) was almost totally from the large
tank. Based on the results of the study, it appeared
that the contamination had been occurring over a
7-3
-------
A Reference Guide for Utilities
period of time, which is consistent with the possibil-
ity of bird contamination. It is likely that the con-
taminant was pulled through the system during the
flushing program. The application of EPANET to the
outbreak proved to be a vital part of the study.
7.2 Reconstructing Historical
Contamination Events - Dover
Township (Toms River), NJ
This case study is focused on evaluating the distribu-
tion of chemically contaminated source water in a
distribution system. The supporting investigations
for this case study were primarily sponsored by the
ATSDR. The investigations involved several other
organizations. The major contributors included the
New Jersey Department of Health and Senior Services
(NJDHSS), the Multimedia Environmental Simula-
tions Laboratory at the Georgia Institute of Technol-
ogy, EPA's National Risk Management Research
Laboratory, and the U.S. Geological Survey.
Key Phrases to Characterize Case Study: historical
reconstruction, hydraulic and water quality modeling,
exposure modeling, contamination assessment, source
tracing, source contribution, model calibration,
sensitivity analysis, genetic algorithm.
7.2.1 Case Study Overview
In August 1995, responding to an evaluation re-
quested by the ATSDR, the New Jersey Department of
Health (now NJDHSS) determined that the childhood
cancer incidence rate in Dover Township (and the
Toms River section) was higher than expected for all
malignant cancers combined (brain and central
nervous system cancer, and leukemia, Berry, 1995). In
March 1996, NJDHSS and ATSDR developed a Public
Health Response Plan (PHRP) describing actions
these agencies would take to investigate the unex-
pected increase in childhood cancers and environmen-
tal concerns in Dover Township (NJDHSS and
ATSDR, 1996). The PHRP included a list of several
evaluations. One of the key evaluations was to
identify potential environmental exposure pathways
relative to two National Priorities List (NPL) sites in
Dover Township (Figure 7-2) - Ciba-Geigy and Reich
Farm. Figure 7-2 also shows the two public water
supply well fields (Parkway and Holly) that were
identified as potential routes of exposure. These well
fields are not only located in the vicinity of the
aforementioned NPL sites, but are also in areas where
the statistically higher childhood cancer rates were
established.
The ensuing evaluations revealed the presence of a
previously unidentified compound, styrene acryloni-
trile (SAN), in the groundwater from the Parkway well-
field that could be traced to the Reich Farm NPL site.
Roads, hydrogiaphy. and boundanes based on 1395 TIGERTUne dala
EXPLANATION
NEW JERSEY
I 1 Reich Farm NPL Site
I I Ciba-Qeigy NPL Site
I I Dover Township Municipal Landfill
I I Dover Township
1 I Water body
• Municipal well
A Storage tank
— Water pipeline
Major road
Hydrography
Figure 7-2. Investigation Area, Dover Township, Ocean
County, NJ (modified from Maslia et a/., 2001).
Similarly, a search of historical records revealed
contamination (primarily semivolatile organics
[SVOCs]) of the Holly well fields that could be
traced to the Ciba-Geigy NPL site. Furthermore,
one of the hypotheses for the epidemiologic case-
control study was that the higher cancer incident
rate was related to the higher exposure to public
water supplies with documented contamination (the
Parkway and Holly well fields). To assist NJDHSS
with the contaminated drinking water exposure
assessment component of the epidemiologic study,
ATSDR developed a water distribution model for
the study area using the EPANET software. This
network model was used to simulate historical
characteristics of the water distribution system
serving Dover Township from 1962-1996. Because
there was a lack of historical contaminant-specific
data during most of the period relevant to the
epidemiologic study, the modeling effort focused
on estimating the percentage of water that a study
subject might have received from each well that
supplied water to the impacted area. The following
subsections present a brief overview of the water
distribution modeling effort (both hydraulic and
water quality) followed by a summary of findings
and conclusions.
7-4
-------
A Reference Guide for Utilities
Prior to the ATSDR's analysis of well field contamina-
tion in Dover Township and the potential linkages to
childhood diseases, another study in Woburn, Massa-
chusetts, heralded the era of such analyses. Though
both the model and graphical presentations are primi-
tive by today's standards, they were effective in provid-
ing a quantitative basis for assessing the spread of
contaminants in the distribution system. The following
is a brief description of the Woburn analysis.
In May 1979, the Massachusetts Department of Environ-
mental Quality Engineering discovered that two wells
(Wells G & H in the B Zone - See Figure 7-3) in
Woburn, Massachusetts were contaminated with toxic
chemicals. Subsequent analysis showed that parts of the
city experienced elevated levels of childhood leukemia
and other illnesses attributed to drinking water derived
from these wells. This event resulted in legal action, a
diverse set of scientific studies that are still ongoing,
and the publication of a book entitled A Civil Action
(Harr, 1995). Early steady-state distribution system
hydraulic and water quality models were also used as a
means to track the movement of the contaminated water
in the distribution system under a range of operating
and demand conditions (Murphy, 1986). The accompa-
nying figure is one example of a plot resulting from this
early modeling effort. As shown in Figure 7-3, based on
the modeling, the city was divided into three zones for
each scenario - the A zone that received no water from
the contaminated wells, the B zone that received all of
its water from the contaminated wells, and the C zone
that received some of its water from the contaminated
wells.
Zion Hill Tank
Figure 7-3. Distribution System Zones - Woburn, MA
(May 1969).
7.2.2 Overall Modeling Approach
Because of the lack of historical hydraulic and water
quality information, the water distribution system was
characterized using data gathered during an extensive
field investigation in 1998. The 1998 field investiga-
tion consisted of two components: (1) determining
spatial locations of distribution system facilities
(wells, tanks, pump, and hydrants) and (2) equipping
hydrants with continuous-recording digital data
loggers and monitoring supply sources (wells, pumps,
and tanks) to measure system responses during winter
demand (March) and summer demand (August)
periods. Twenty-five hydrants located throughout the
distribution system were equipped with data loggers
to simultaneously collect information on system
response (Maslia et al., 2000). The collected response
data included on-off cycling of groundwater wells,
high service and booster pump operations, pressure
variations, storage tank water-level fluctuations, and
total production.
A detailed "all-pipe" hydraulic network model was
developed and calibrated to present-day conditions
(1998) using the field investigation results. The
reliability of the calibrated model was successfully
demonstrated through a water quality simulation of
the transport of a naturally occurring conservative
element (barium) and a comparison of the results with
data collected in March and April 1996 at 21 schools
and 6 points of entry to the water distribution system.
Thereafter, to describe the historical distribution
system networks specific to the Dover Township area,
databases were developed from diverse sources of
information. These data sources included water utility
pipeline installation records, quarterly billing records,
NJDHSS groundwater well records, and annual water
utility reports to the state board of public utilities.
These data were applied to EPANET and simulations
were conducted for each month of the historical
period—January 1962 through December 1996 (420
simulations). After completing those 35-year/420-
month analyses, source-trace analysis simulations
were conducted to determine the percentage of water
contributed by each well or well field operating
during each month for all study subject locations.
A review of the historical network configuration
revealed that the water distribution system complex-
ity increased significantly during this period. The
model inputs were appropriately adjusted to account
for these historical changes. For example, the 1962
water distribution system was represented with an
approximate peak production of 1.3 million
gallons per day (MGD) produced from three wells that
served nearly 4,300 customers (population -17,200).
By contrast, in 1996, the water distribution system
had an approximate peak production of 13.9 MGD
produced from 20 wells that served nearly 44,000
7-5
-------
A Reference Guide for Utilities
customers (population ~ 89,300). Appropriate
adjustments were made to modeled pipe segments,
storage reservoirs, and operational details. Grayman
al. (2004) present a more detailed
account of the EPANET model
input adjustments. Produc-
tion data for the 420-
month historical 1988
period is graphi-
cally represented
in Figure 7-4.
et
1995
I ,
Figure 7-4. Three-Dimensional Representation of Monthly
Production of Water, Dover Township Area, NJ (from
Maslia et al., 2001).
To perform an extended period simulation (EPS) of
the distribution of water for each of the 420 months of
the historical period, information was required on
network configuration, demand, and operational data.
However, operational data prior to 1978 were unavail-
able, requiring the development of system operation
parameters—designated as "master operating criteria
(MOC)." The MOC is based on hydraulic engineering
principles necessary to successfully operate distribu-
tion systems similar to the one serving the Dover
Township area (Table 7-1). From 1978 forward, for
selected years, operators of the water utility provided
information on the generalized operating practices for
a typical peak-demand (summer) and non-peak
demand (fall) day. These guidelines were used in
conjunction with the MOC to simulate a typical 24-
hour daily operation of the water distribution system
for each month of the historical period.
The model parameter of interest from the epidemio-
logic study perspective was the proportionate
contribution of water from wells and well fields to
locations throughout the historical pipeline networks.
Thus, the distribution of water delivered to pipeline
locations was the item of interest rather than the
specific operations of the wells, storage tanks, and
pumps (WSTP) that delivered the water. Normally,
detailed WSTP operational inputs would be required
for EPANET simulation. However, to simplify the
simulation methodology and reduce data require-
ments, a "supply-node-link" (SNL) method of
idealizing the WSTP combination was developed. In
the SNL simulation method, an equivalent amount of
water is supplied to the distribution system (based on
estimated monthly demands and the typical daily
operation of the systems). To demonstrate that the
surrogate SNL simulation method supplies the
distribution system with an equivalent amount of
water when compared to the real-world WSTP
simulation method, both simulation methods were
applied to the present-day (1998) water distribution
system for conditions existing in August 1998. The
results obtained from these simulations produced
nearly identical flows in the modeled system.
7.2.3 Simulation Techniques
Using the EPANET network model developed for the
Dover Township area, hydraulic modeling was
conducted whereby average network conditions were
simulated for every month of the historical period
Table 7-1. Master Operating Criteria Used to Develop
Operating Schedules for the Historical Water Distribution
System, Dover Township Area, NJ (from Maslia et al.,
2001)
Parameter
Pressure1
Water level
Hydraulic device
on-line date
On-and-off cycling:
Manual operation
On-and-off cycling:
Automatic
operation
Operating hours
Criteria
Minimum of 15 psi; maximum of 1 10
psi at pipeline locations, including
network end points
Minimum of 3 ft above bottom
elevation of tank; maximum equal to
elevation of top of tank; ending
water level should equal the starting
water level
June 1 of year installed to meet
maximum-demand conditions
Wells and high-service and booster
pumps cannot be cycled on-and-off
from 2200 to 0600 hours
Wells and high-service and booster
pumps can be cycled on-and-off at
any hour
Wells should be operated
continuously for the total number of
production hours, based on
production data2
'Generally, for residential demand, minimum recommended
pressure is about 20 psi. However, for some locations in
the Dover Township area (mostly in areas near the end of
distribution lines), lower pressures were simulated.
2See Maslia et al. (2001) for production data (Appendix B)
and hours of operation (Appendix D)
7-6
-------
A Reference Guide for Utilities
(420 simulations). These simulations were completed
under balanced flow conditions that utilized hydrau-
lic engineering principles and conformed to the MOC
(Table 7-1). Thereafter, using the results of the
monthly network hydraulic simulations, water quality
simulations (source-trace analysis) were conducted for
each water source (point of entry) of the network in
order to determine the monthly proportionate
contribution of source water at all locations in the
Dover Township area serviced by the water distribu-
tion system.
EPANET is a dynamic water quality model that has
the ability to compute the percentage of water
reaching any point in the distribution system over
time from a specified location (source) in the
network. To estimate this proportionate contribu-
tion of water, a source location is assigned a value
of 100 percent. The resulting solution provided by
the water quality simulator in EPANET then
becomes the percentage of flow at any location in
the distribution system network (for example, a
demand node) contributed by the source location of
interest. For the purposes of this analysis, a source-
trace analysis was conducted for every month of the
historical period. Source nodes were assigned a
value of 100 percent in order to estimate the
proportionate contribution of water to locations in
the historical distribution system networks. These
initial conditions were fully propagated through
most of the distribution system before retrieving
the proportionate contribution results (Maslia et al.,
2000). Accordingly, for each monthly historical
network model, 24-hour demand and operational
patterns were defined and these patterns were
repeated for approximately 1,200 hours to reach a
state of stationary water-quality dynamics (dynamic
equilibrium). For most of the analyses, hydraulic
time steps of 1 hour and water-quality time steps of
5 minutes were used within EPANET. For some
monthly simulations, the water-quality time steps
were reduced to 1 minute to ensure that the mass
balance summed to -100 percent (range of 98 to
101 percent due to numerical approximations).
With respect to the scheduling of groundwater well
operations, the EPANET model was set to utilize
pattern factors corresponding to the hourly operations
of supply wells. These pattern factors along with the
operational extremes of storage tank water levels were
manually adjusted during each of the 420 monthly
network simulations to achieve balanced flow
conditions. This approach to simulation was desig-
nated as the manual adjustment process. A second
simulation approach designated as the genetic
algorithm (GA) approach was also utilized to achieve
balanced flow conditions for each of the 420 monthly
networks of the historical period. This approach
required the development of an innovative methodol-
ogy known as the progressive optimality genetic
algorithm (POGA), which is an automated objective
simulation technique (Aral et al., 2004a, b). The GA
simulations utilized the balanced flow conditions
obtained by the manual adjustment process as starting
conditions. The GA technique was used to address
the following key questions:
• If a balanced flow operating condition was
achieved using the manual adjustment process,
was the resulting operating condition the only
way the system could have been successfully
operated?
• Could alternative or additional operating
conditions be defined such that system
operations would also be satisfactory or even
optimal?
Thus, the POGA methodology was used in conjunc-
tion with EPANET to simulate alternative and
possibly optimal water distribution system operations
and to assess the effects of variations in system
operations on the results of the proportionate contri-
bution simulations.
7.2.4 Simulation Results and Conclusions
Figure 7-5 shows the aerial distribution of simulated
proportionate contribution results for all model nodes
(pipeline junctions) for the month of July 1988, using
the Parkway well field as the point of entry (source
point). The simulated proportionate contribution
results are divided into six intervals (1 to 10 percent,
10 to 25 percent, 25 to 50 percent, 50 to 75 percent,
75 to 90 percent, and 90 to 100 percent) and a color is
assigned to all nodes within each interval (results are
not shown for negligible proportionate contributions
of less than 1 percent).
Simulated proportionate contribution results can also
be viewed in terms of selected pipeline locations and
the combination of wells or well fields that contribute
water to that location. Five geographically distinct
pipeline locations are selected from the historical
networks to represent the spatial distribution of
proportionate contribution results. These locations
are identified on Figure 7-5 as locations A, B, C, D,
and E. The simulated proportionate contribution of
water for July 1988 corresponding to each pipeline
location is shown graphically on Figure 7-6. The
simulation results demonstrated that the contribution
of water from wells and well fields varied by time and
location. However, the results also showed that certain
wells provided the predominant amount of water to
locations throughout the Dover Township area.
Additionally, although the pattern factors for some
hours of operations showed marked differences, the
simulated proportionate contributions of water using
7-7
-------
A Reference Guide for Utilities
, South Toms River
• IIVIUO
•Wells 32.38
CD
cn
dl
EXPLANATION
Reich Farm NPL Site ® Municipal well
A Storage lank
Ciba-Geigy NPL Site
Dover Township
Water body
Water pipeline
Major road
Hydrography
I Pipeline location and letter—Percent contribution is report
Percentage of water contributed by Parkway wells
(22. 23. 24, 26. 28. 29). 24-hour average
1 to 10 25 lo 50 • 75 to 90
101025 501075 • 9010100
Notes, in Wale* pipelines lange in diameter ttom 2 inches lo 16 inches 13) Pipeline Irom walof-ulrlity database (Flegal 1937)
(2) Roads, hydrography, and boundaries based on 1995 (4) Percentage 01 water based on model reaching
TIOER/Une data dynamic quillbrium after 1.200 hours ot simulation
Figure 7-5. Areal Distribution of Simulated Proportionate
Contribution of Water from the Parkway Wells (22, 23, 24,
26, 28, 29) to Locations in the Dover Township Area, NJ,
July 1988 Conditions (from Maslia et al., 2001).
O 100
SJ90
_j 70
UJ
I 60
o
E 50
A B C D E
SELECTED PIPELINE LOCATION
Well or well field—Number in
parenthesis is well number.
See Fiyure 7.3 lur pipeline lutdliun
Holly (21.30)
Brookside(15)
South Toms River (32.38)
Indian I lead (20)
Parkway (22.23.24.26.28.29)
Route 70 (31)
Berkeley (33.34.35)
Figure 7-6. Simulated Proportionate Contribution of Water
from Wells and Well Fields to Selected Locations, Dover
Township Area, NJ, July 1988 Conditions (from Maslia et
al., 2001).
pattern factors derived from the application of the
POGA methodology showed little difference through-
out the Dover Township area when compared to
corresponding proportionate contribution of water
simulated using the manual adjustment process. The
results of sensitivity analyses conducted using the
historical reconstruction process indicated the
following:
• There was a narrow range within which the
historical water distribution systems could
have successfully operated and still satisfy
hydraulic engineering principles and the
MOC.
• Daily operational variations over a month did
not appreciably change the proportionate
contribution of water from specific sources.
Therefore, the reconstructed historical water
distribution systems were determined to be the most
plausible and realistic scenarios under which the
1962-1996 historical water distribution systems
were operated. The health scientists conducting
the case-control epidemiologic study used the
resulting percentage of water derived from the
different sources to derive exposure indices for
each study subject.
The results from the case-control study showed that
there was an association between prenatal exposure to
contaminated community water and leukemia in
female children (NJDHSS, 2003). For example, female
leukemia cases were 5 times more likely to have
occurred when exposed during the prenatal period to
a high percentage of Parkway well water than were
control children. The control children are those
living in the study area, but were not exposed to the
water from the contaminated well fields. These
findings would not have been possible without the
results derived from the innovative water distribution
system modeling efforts. These efforts have led to
developing new methods for evaluating the accuracy
of modeling results and exposure classification
techniques that are critical components of epidemio-
logic studies. Some of the innovations documented
by the Dover Township historical reconstruction
analysis are:
• A new approach, proportionate contribution
analysis, was developed that utilized water
distribution system modeling and source
tracing to quantify exposure on a monthly basis
for all locations historically served by the
distribution system.
• Through the use of an innovative genetic
algorithm approach (POGA), historical water
distribution system operating schedules were
7-8
-------
A Reference Guide for Utilities
The city of Redlands lies in the San Bernardino valley of California, approximately 60 miles east of central Los
Angeles. In 1981, a routine analysis for chlorination byproducts revealed the presence of trichloroethylene
(TCE) in a sample of water from the Redlands water system. Subsequent water quality analyses revealed that a
number of wells supplying the city were contaminated with TCE. In 1997, the perchlorate anion (ClO^ was
also detected in several wells. In 1996, the first of a series of lawsuits was filed in California State Court
alleging that the source of these contaminants was a manufacturing facility located up-gradient from the most
seriously contaminated wells.
One of these lawsuits claimed that plaintiffs were harmed by exposure to toxic chemicals that were improperly
disposed of at the manufacturing site and found their way into groundwater that was subsequently extracted
through the city's wells and delivered to water customers, including the plaintiffs. The plaintiffs' burden of
proof requires them to establish, among other things, that they were actually exposed to contaminated water at
their homes, places of work, or other locations and that the amounts of contaminants that entered their bodies
as the result of these exposures were sufficient to cause harm to them.
To establish this proof, experts for the plaintiffs
reconstructed the historical conditions in the water
distribution system of the City of Redlands over a
period from the mid 1950s to the late 1990s using
the EPANET model. As part of litigation, several
forensic reconstructions of water quality in the
Redlands water distribution system were performed.
The reconstruction involved estimates of both
human exposure to toxic contaminants and whole-
body intakes of these chemicals. Estimates of
exposures and intakes were expressed as credibility
intervals, which were calculated using Monte Carlo
simulation techniques. As an example output from
the analysis, Figure 7-7 illustrates the estimated
upper 97.5 percent credibility limit of one
plaintiff's exposure to perchlorate. Similar informa-
tion was developed to describe the lower 2.5
percent credibility limit for each of the test plain-
tiffs in the case (Grayman et al., 2004).
Individual 3
Year
Figure 7-7. Estimated Upper 97.5 Percent Credibility
Limit for Annual Perchlorate Intake by One Plaintiff
(Grayman, 2004).
synthesized. Sensitivity analyses indicated
that operating system changes did not
appreciably change the proportionate
contribution of water to Dover Township
locations.
• The association between exposure and disease
would not have been possible without
developing the integrated approach using
environmental science, engineering
evaluations, and epidemiologic analyses.
Historical reconstruction of environmental exposure is
not an easy task. The procedures and results summa-
rized herein (and the detailed analyses in Maslia et
al., [2001]) represent one of the most comprehensive,
well-documented, and quality-controlled studies of its
kind. Another example of public exposure assessment
using modeling is related to the City of Redlands,
California.
7.3 Application of Water
Distribution System Modeling
in Support of a Regulatory
Requirement
The new DBPR2 regulation that is proposed for
promulgation in the near future requires all water
utilities that have a disinfectant residual in the
distribution system to perform an IDSE unless they
obtain a small-system or "40/30" waiver (EPA, 2003).
Systems that can certify TTHM and HAAS compli-
ance data to be less than or equal to 40 |ig/L for
TTHM and 30 \ig/L for HAAS are not required to
perform an IDSE. The goal of the IDSE is to identify
compliance monitoring sites that may have high DBF
levels in distribution systems. Utilities may choose to
perform an SMP that involves extensive monitoring.
Alternatively they may choose to perform a system-
7-9
-------
A Reference Guide for Utilities
specific study (SSS) that uses historical data, distribu-
tion system models, or other analyses combined with
minimal monitoring to evaluate TTHM and HAAS
levels throughout the distribution system as the basis
to select future compliance monitoring sites. This
case study demonstrates how a hydraulic/water
quality distribution system model can be applied to
satisfy the IDSE requirements of an SSS.
Key Phrases to Characterize Case Study: Regula-
tory modeling, IDSE, water quality modeling.
7.3.1 IDSE Requirements Overview
The IDSE guidance manual spells out a series of
suggested minimum requirements for the use of a
calibrated water distribution system hydraulic model
to perform an SSS. In general, the water distribution
system model should be more comprehensive for the
purpose of an SSS than models typically used for
long-range capital improvement program analysis
(e.g., master planning). A calibrated hydraulic model
intended for detailed distribution system design (e.g.,
for system improvements) or operational studies is
likely to be adequate. Because systems are always
changing (e.g., population growth, industry develop-
ment in network area, aging of mains), it is important
that the model generally reflect system conditions and
demand at the time of the IDSE SSS. A model that has
not been updated or calibrated in the last 5 to 10
years is unlikely to be adequate for an SSS. The
guidelines provided in the draft guidance manual are
summarized below:
• EPS model that has been recently calibrated
using generally accepted methods.
• An all-pipe model or skeletonized model that
includes (a) at least 50 percent of total pipe
length in the distribution system, (b) at least 75
percent of the pipe volume in the distribution
system, (c) all 12-inch-diameter and larger
pipes, (d) all 8-inch and larger pipes that
connect major facilities, (e) all 6-inch and larger
pipes that connect remote areas of a distribution
system, and (f) all active control valves or other
system features that could significantly affect
the flow of water through the distribution
system.
• Water demand should be allocated among the
nodes of the model in a manner that reflects the
actual spatial distribution of such demand
throughout the system.
• A system-specific, diurnal (24-hour) demand
pattern should be applied to the overall system
demand.
• The model should accurately simulate seasonal
system configurations and operational changes.
Once the model is established, it is then run in EPS
mode until a consistent, repeating temporal pattern of
water age is established at all nodes of the model.
Generally, the model should be run under high DBF
formation conditions (typically summer months) to
estimate residence times. Based on the modeled water
age results, preliminary monitoring sites are identified
near locations that satisfy the sampling site require-
ments. Sampling sites are selected to represent:
• High-TTHM Sites: High TTHM values are
expected at high-residence-time locations.
These locations can be identified by reviewing
the modeled water age at each node in the
model. These sites are generally downstream of
storage facilities and in remote locations.
However, the regulation does not require
extremes or non-representative sites to be
sampled.
• High-HAA5 Sites: The criteria and procedure
for selecting high HAAS sites using a hydraulic
model are generally the same as those described
above for selecting high-TTHM sites with one
important difference: the sites chosen to
represent high HAAS should have a disinfectant
residual sufficient to suppress bacteria which
can degrade HAAs.
• Average-Residence-Time Sites: Average-
residence-time sites can be selected from sites
with residence times close to the flow-weighted
mean of all nodal residence times.
• Near-Entry-Point Sites: Modeled water age can
be used to identify locations in the near vicinity
to entry points into the water system.
Requirements for the number of monitoring sites have
not yet been finalized. As a result, this case study
demonstrates the general usage of models for IDSE
and relies upon the 2003 draft guidance issued by
EPA to illustrate the usage of models.
7.3.2 Example Application of Modeling in the
IDSE Process
The following example is a hypothetical case study
based in large part on an actual water distribution
system. The system purchases disinfected groundwa-
ter and serves approximately 15,000 people. Water
enters the distribution system from two separate
interconnections to a wholesale utility. The average
demand is 2.2 MGD. A north interconnection operates
intermittently and provides approximately 80 percent
of the demand, while the south interconnection
operates at all times and provides the remaining 20
percent of the system demand. There is a 1.5-million-
gallon storage tank. In order to comply with the Stage
2 requirements, the draft proposed DBPR2 states that
a total of six sites are required for a utility of this size
7-10
-------
A Reference Guide for Utilities
using groundwater: one representing a site near to the
predominant entry, one representing a site with
average residence time, two sites representing high
TTHM conditions, and two representing high HAAS
conditions.
The water utility water distribution system model has
been used extensively in the past for both hydraulic
and water-quality studies. It is a skeletonized model
that includes all 8-inch-diameter and larger pipes and
all major facilities. The pipes in the model represent
74 percent of the total length of pipe in the system
and 86 percent of the total volume. The model has
been previously calibrated based on two tracer studies
and has been shown to have excellent agreement with
observed field results. Demands have been assigned
to nodes based on actual meter readings, and informa-
tion from the SCADA system has been used to construct
a typical diurnal water use pattern. The model is being
operated in an EPS mode to simulate a 12-day period.
The model has also been calibrated for use in simulat-
ing chlorine residual in the distribution system.
A series of runs of the model were performed to help
understand the movement of water and water quality
transformations in the system. Specific simulations
included water age and chlorine residual. The results
of the water-age model run are shown in Figure 7-8.
This plot shows the average water age throughout the
distribution system over the last 24 hours of the 2-
week simulation. This period was selected to avoid
the uncertainty associated with assigning initial water
age in the system. The plot illustrates the nodes in the
vicinity of the predominant northern interconnection
that receive water with an average age of less than 2
hours. As also shown, the average water age increases
significantly for areas that are further from the
interconnections. The demand flow-weighted average
water age for delivered water was calculated to be 27
hours. However, water age can also vary quite
significantly over the course of a day in water
systems, largely due to the impacts of storage tanks.
This is illustrated in Figure 7-9, which depicts the
minimum water age at each node over the same 24-
hour period. This is also shown in the plot in Figure
7-10 for Node J-456 in the vicinity of the tank. The
IDSE guidance does not require utilities to explicitly
consider the effects of tanks on diurnal variations in
water age, and thus on the formation of DBFs. If Node
J-456 was selected as representative of high DBF
because of its high water age, it would be expected
that the DBFs would only be high during the part of
the day when water is being discharged from the tank.
The model was also used to determine the chlorine
residual throughout the system. Figure 7-11 contains
a plot of the minimum chlorine residual throughout
the system. It is important to note areas with high
North Interconnection
Figure 7-8. Average Water Age in the Distribution
System Over Last 24 Hours of a 2-Week Simulation.
North Well
Figure 7-9. Minimum Water Age in the Distribution System
Over Last 24 Hours of a 2-Week Simulation.
7-11
-------
A Reference Guide for Utilities
Node J-456
150.0
Age for Node J-456
125.0
100.0
75.0
50.0
25.0
0.0
A
V
264 268 272 276 280 284
Time (hours)
288
Figure 7-10. Diurnal Water Age at Node J-456.
residence time and low residual for high-TTHM sites
and high residence time and high residual for high-
HAA5 sites. This information can be used to avoid
selecting monitoring sites that are to be used as
representative of high HAAS concentrations. The
current ability to accurately model HAAS in a
distribution system is limited. However, research has
shown that depressed chlorine residual can result in
biodegradation of HAAS, thus lowering the HAAS
concentrations even for older water. Figure 7-11 also
shows a small area in the southwestern portion of the
system that the model predicts to potentially experi-
ence chlorine residuals less than 0.2 mg/L of chlorine.
Based on the modeling results, various zones were
defined in the distribution system representing areas
that are appropriate for different types of compliance
monitoring requirements (Figure 7-12). In actual use,
information generated by the model would be
supplemented by historical field data. The zones
shown in the plot include:
1. Nodes in the vicinity of the predominant north
connection with water age less than 2 hours
representative of near entry locations;
2. Nodes with average water age in the range of 21
to 33 hours that represent locations that
approximate the average residence time of 27
hours;
North Interconnection
Figure 7-11. Minimum Chlorine Residual in
Distribution System Over Last 24 Hours of a 2-Week
Simulation.
rconnection
Near Entry Point
Average Residence Time
HighTTHM
High HAA
South Interconnection
Figure 7-12. Zones Representing Potential
Monitoring Locations for IDSE Based on Modeling.
7-12
-------
A Reference Guide for Utilities
3. Nodes with residence times that exceed twice
the average water age (> 54 hours) and have a
minimum chlorine residual exceeding 0.20 mg/L
representing potential high-HAA5 sites; and
4. Other nodes with residence times that exceed
twice the average water age (> 54 hours)
representing potential high-TTHM sites.
As illustrated by the case example, if a detailed,
calibrated EPS model is available, the model repre-
sents an efficient means of defining the compliance
monitor locations as required under the forthcoming
regulation.
7.4 Use of Water Distribution
System Models in the
Placement of Monitors to
Detect Intentional
Contamination
The increasing concern over the potential for inten-
tional contamination of a water distribution system
has led to interest in the placement of monitors to
detect contamination and to serve as part of a rapid
detection system. Design of such monitoring systems
must include decisions on the type, number and
location for the monitors. Water distribution system
models can play a significant role in the decision
making by providing a quantitative mechanism for
determining the movement of a contaminant through
the distribution system and testing the effectiveness
of a monitoring system design.
To illustrate this application, a red team-blue team
concept is used (Grayman et al., 2005). The red
team-blue team concept is part of "war gaming"
that is widely used today as a mechanism for
training and development and testing of security
plans. The red team acts as the aggressor and the
blue team acts as the defenders. Each team has
different types and amounts of information avail-
able to them and different rules or constraints that
they must follow. In this case study, network
models are used in two modes to assist in evaluat-
ing monitoring networks:
1. As part of a red team-blue team exercise to
demonstrate the effectiveness of manual
selection of location of monitors as part of a
CWS.
2. As part of an optimization model to select the
best locations for monitors based on a stated
metric for measuring the effectiveness of the
monitoring system.
Key Phrases to Characterize Case Study: Water
security, contamination, optimization, monitor
placement
7.4.1 Red Team-Blue Team Exercise
In this simulated exercise, the red team attacks a water
distribution system by adding a harmful chemical to
the water. The red team is provided with limited
information on the distribution system, a number of
potential locations where they can inject a contami-
nant, and predetermined information on the character-
istics of the contaminant (quantity and lethality of the
contaminant). The blue team represents the water
utility and attempts to protect the water system by
installing three monitors as part of a CWS that detects
contaminants. It is assumed that they have extensive
information on the design and operation of the
distribution system but no firm information on where
the attackers may choose to introduce the contami-
nant or the nature of the contamination scenario.
The water distribution system network used in this
exercise is a skeletonized version of a major pressure
zone of a water distribution system in California
approximating the conditions (design and operation)
in the mid 1990s. The system is fed by two sources;
one that operates continuously and one that operates
only during the day. There are three storage facilities
located in the network. This network is one of the
example networks provided as part of the EPANET
model. The simulation performed in the exercise is a
24-hour EPS starting at 7AM. The model representa-
tion of the network is shown in Figure 7-13. This
figure also illustrates the relative nodal demands, and
the typical flow directions and magnitude during the
day. This plot is given only to the blue team to
provide them with information on the design and
operation of the system. The red team is provided
Figure 7-13. Water Distribution System Characteristics.
7-13
-------
A Reference Guide for Utilities
• Allowable Contaminant
Introduction Location
O Monitoring Nodes
Related to Figures 7-15 and 7-16
Figure 7-14. Allowable Contaminant Introduction Locations.
only with a map of the distribution system showing
eight potential sites that can be used to introduce a
contaminant (Figure 7-14).
Following the selection of points of attack by the red
team and selection of monitor locations by the blue
team, the contaminant introduction is simulated using
the EPANET model. The movement and concentration
of the contaminants are then viewed graphically by
animating the movement of the contaminant in the
distribution system in the EPANET model. The time
history of contaminant concentrations is also viewed
at selected nodes in the distribution system. The
effectiveness of monitors is illustrated by graphing
the concentrations of the contaminants at monitoring
nodes and assessing whether (and how quickly) the
monitors will serve their purpose of detecting the
contaminant. Figures 7-15 and 7-16 show the
concentrations resulting from 8 hours of contamina-
tion at node 123 starting at 11 a.m.
In Figure 7-15, the resulting contamination at a
node located 1500 feet immediately south of the
injection point is shown. As expected, the contami-
nant moved very rapidly and reached this node in less
than an hour after the injection. If a monitor was
located at this node, and rapid analysis and re-
sponse occurred, it could be very effective as an
early warning for most of the distribution system.
Figure 7-16 illustrates concentrations resulting
from the same contaminant introduction location.
This node is located near the eastern edge of the
distribution system approximately 2.5 miles
downstream of the contaminant introduction
location. As illustrated, the concentration of the
contaminant remained about the same but the travel
time to this point was approximately 7 hours. For
this injection scenario, a monitor located at this
point would be relatively ineffective as a warning
device for most of the distribution system because
of the significant time lag.
In the exercise, most red team members tend to select
contamination introduction locations that they
perceive would result in the most widespread impacts.
The most often selected sites were those near to the
water sources. Little attention is generally given to
the amount of dilution that would result at a particu-
lar location. Blue team members tend to select
monitoring locations that cover a wide range of
locations. Frequently, the three allowable monitors
will be located in the north, central, and southern
portions of the distribution system.
7.4.2 Application of Optimization Model
The optimization model used in this demonstration
is a methodology developed by Ostfeld and
Salomons (2004). The model links EPANET and a
genetic algorithm in an overall framework for
optimally allocating monitoring stations, aimed at
detecting deliberate external contamination into
water distribution system nodes. The model
operates under extended period (unsteady) hydrau-
lics and water quality conditions. The optimization
routine determines the monitor placement to detect
contaminants in order to minimize the exposure of
Chemical for Node 121
10 12 14
Time (hours after 7 AM)
Figure 7-15. Contaminant Concentration Just Downstream
of Contaminant Introduction Location (Node 121).
45.0
40.0
35.0
=730.0
"S
.§25.0
| 20.0
£ 15.0
o
10.0
5.0
0.0
Chemical for Node 143
10 12 14 16 18 20 22 24
Time (hours after? AM)
Figure 7-16. Contaminant Concentration Far Downstream
of Contaminant Introduction Location (Node 143).
7-14
-------
A Reference Guide for Utilities
customers above an allowable minimum concentra-
tion. The algorithm can be used to study contami-
nation of fixed duration, quantity, or location or
can simulate contamination under stochastic
conditions. There are several model parameters that
can be specified to control the number of monitors,
the allowable contaminant introduction locations, the
characteristics of the event, and whether the event
characteristics and demands are to be considered as
stochastic variables.
In one application of the model, it was assumed that
the pollutant could be introduced at any single node
of the distribution system at any time, all with the
same injection probability. The following additional
assumptions were made:
• The maximum contamination exposure volume
to the public above which an alarm signal of the
monitoring stations is required is 25 gallons.
• The water is considered contaminated above 1
mg/L.
• The pollutant flow discharge is 2 kg/min for 5
minutes (i.e., a total of 10 kg of a solution of 100
percent is introduced within a total of 5 minutes).
• The pollutant flow discharge of the
contaminant introduced and the consumer
demands are deterministic.
• The monitoring stations are providing real-time
data and detection alarms.
• All monitoring stations have a detection
sensitivity of 1 mg/L.
• 3 monitors are to be placed.
The model suggests placing monitors at nodes 143,
181, and 213 with a detection likelihood of 0.4354
(i.e., there is a probability of about 44 percent that the
contaminant will be detected prior to the consump-
tion of more than 25 gallons at a concentration higher
than 1 mg/L). The location of the monitors is shown
in Figure 7-17. As illustrated, the selected monitor
locations were relatively evenly spaced around the
network.
Other evaluated scenarios looked at a different
number of allowable monitors, the allowable contami-
nant introduction locations, the critical exposure
threshold, and representation of contaminant quantity
and nodal demands as stochastic variables. Though
the exact "optimal" locations varied slightly between
the different runs, typically the monitors were placed
throughout the network. However, the effectiveness of
the monitoring network, as measured by the detection
likelihood does vary considerably between scenarios.
Node 143
Node 181
Figure 7-17. Monitoring Locations Selected by the
Optimization Model.
7.4.3 Case Summary
The red team - blue team exercise serves as a good
mechanism for demonstrating both the dynamics of
contaminant movement in the distribution system and
the potential effectiveness of monitors. Application of
the optimization model, both as a demonstration
procedure and as a practical tool, provides a method
that moves the monitor placement from a purely
intuitive process to a quantitative procedure. Both the
exercise and the optimization tool show the impor-
tance in minimizing delays in the detection, notifica-
tion, and response process. Even an added delay of an
hour or two can lead to a significant increase in the
number of customers that would be impacted by a
contamination event.
7.5 Case Study - Use of
PipelineNet Model
This case study focuses on the application of the
PipelineNet model, which incorporates both GIS and
the EPANET model discussed in the previous chapters
of this reference guide. The supporting investigations
were primarily sponsored by the Awwa Research
Foundation (AwwaRF) and EPA with work performed
by a consulting firm (SAIC) and assistance from water
utility personnel (Ron Hunsinger, Bill Kirkpatrick,
Dave Rehnstrom) working at the East Bay Municipal
Utility District (EBMUD), Oakland, CA. The text and
figures are adapted from AwwaRF report 2922
prepared by Bahadur et al. (2003).
Key Phrases to Characterize Case Study: hydraulic
and water quality modeling, placement of monitors,
exposure modeling, contamination assessment,
contamination response tools, geospatial analysis.
7-15
-------
A Reference Guide for Utilities
7.5.1 Overview
PipelineNet is an EPANET/ArcView1 -based model,
and uses the same hydraulic engine as EPANET. The
EPANET portion of the model can simulate the fate
and transport of potentially introduced contaminants
in a water distribution system. The ArcView (or the
GIS layer) portion of the model can relate the
geospatial components of the simulated impact. The
GIS layer allows for geo-features and map display
with an overlay of model output. This feature is
particularly useful in applications such as emergency
response, determining optimal placement of sampling,
and monitoring instruments.
AwwaRF and EPA jointly funded a project to
develop techniques to locate monitoring points in a
distribution system, determine appropriate timing
and frequency of monitoring, and establish moni-
toring techniques and relevant water quality
parameters. For this purpose, a fully calibrated
extended period simulation (EPS) network model
hypothetically representing a portion of EBMUD
was developed using PipelineNet. This study area
network model represents 16 of the 123 pressure
zones in the overall EBMUD distribution system.
The study area contained 27 tanks, 748 miles of
pipes, 62 pumps and 17,997 pipe segments with
diameters equal to or greater than 2 inches. Figure
7-18 shows a partial view of the hypothetical
network of pipelines.
7.5.2 Model Calibration
The network model was calibrated by comparing
the observed (SCADA data) and simulated
(PipelineNet model) water level in 25 tanks located
in the study area. The primary focus of the calibra-
Figure 7-18. Hypothetical Water Distribution System
Showing Pipelines.
tion was to match the shape of the observed water
level in the tanks. The model calibration was
performed for a 24-hour time interval using data
measured on July 1, 2001. To further enhance
calibration, the pump characteristic curves were
used to achieve a good comparison between
simulated and observed tank levels. The flow value
of the characteristic curve was changed as neces-
sary to reflect field conditions. Each pump was
operated with time controls.
7.5.3 Monitoring Site Location Methodology
A hierarchical selection process was developed to
locate monitoring stations in the distribution system.
A three-step approach was employed based on model
inputs, outputs, and GIS layers (see Figure 7-19).
Hierarchical Selection Concept
StepS
Critical Facilities & Population Density
Step 2
Distribution System Response
Step 1
Source Prioritization
Water Distribution System
Figure 7-19. Conceptual Diagram Showing the Ranking
and Prioritization Methodology.
In the first step (source prioritization factor), all the
elements of the water distribution system are
assumed to be available for monitoring. This
universe is then reduced to a smaller set based on
accessibility (location) and amenability (e.g.,
eliminating dead ends, crosses, tees, junctions with
different pipe material) to monitoring. Initially, all
nodes are considered available for monitoring and
are assigned a score of 1. Subsequently, all the
nodes, which are either not amenable or not
accessible, are assigned a score of 0. This reduced
the number of pipes available for monitoring from
17,997 to 14,938. Therefore, only the 14,938
nodes with a score equal to 1 are considered for
Step 2 (described below).
In the second step (distribution system response
factor), the PipelineNet model is run to quantify the
distribution system response in terms of flow, veloc-
ity, and pressure. Concentration of water quality
parameters could also be considered in this ranking
procedure but was not included as a factor in this case
study. Each system response parameter has equal
'Registered Trademark of ESRI
7-16
-------
A Reference Guide for Utilities
weighting and is assigned an initial score of 1 for
every pipe. Thereafter, based on the run, the scores
are re-assigned values ranging between 1 and 10,
where a score of 10 would indicate a higher level of
concern. For any given parameter, the user can
determine the distribution of scores over the param-
eter range. For example, a score range of 10 to 1
could be distributed over a flow range of 0.001 to 100
gpm. The PipelineNet model and Bahadur et al.
(2003) provide some guidance for assigning scores,
but the user can select any score based on the require-
ments of the analysis.
In the third step (critical facilities and population
density factor), user defined buffer zones (polygons)
are created around critical facility locations. In
addition, areas of low-, medium-, and high-population
density are delineated by the creation of polygons.
Pipes closest to the critical facilities and/or near high
population density areas are assigned a score of 10.
The total score for each pipe based on Steps 2 and 3
are computed. These final scores are linked to the GIS
pipeline layer. The user can identify areas where
monitoring stations should be placed based on the
display of pipes with high scores. Figure 7-20 shows
the pipes in the hypothetical network with scores
greater than 27 overlaid with critical facility loca-
tions. The methodology outlined above for selecting
the location of monitoring stations is a subjective
procedure that requires input and judgment from
water utility personnel. It would not result in a
common solution for all distribution systems but can
incorporate the specific needs and objectives of the
participating water utility.
Figure 7-20. Hypothetical System Showing High Score
Areas (>27) Overlain with Hospitals and Schools.
7.5.4 Response and Mitigation Tools
In addition to the monitoring site location methodol-
ogy, three additional tools were developed as part of
this case study to enhance the capability of
PipelineNet in the areas of emergency response,
mitigation, and normal operations. These three tools
are briefly described in the following subsections.
7.5.4.1 Consequence Assessment Tool
The consequence assessment tool of PipelineNet
provides the ability to quickly identify and quantify
the population, infrastructure, and resources at risk
from a contaminant event. For a defined contami-
nated area, this tool can calculate:
• total population at risk,
• number of taps contaminated,
• miles of pipe contaminated,
• total number of hospitals and beds for each
hospital, and
• total number of schools and student population.
7.5.4.2 Isolation Tool
The isolation tool of PipelineNet provides the ability
to change the status (open or closed) of any pipe in
the distribution system. After completing a water
quality simulation and examining the contaminant
distribution from the event, this tool could be used to
close off one or more pipes to control the flow of
water. The model would then be re-run, reflecting
these new hydraulic conditions, and the output
examined to determine if this mitigation step was
successful in limiting the area of contamination.
7.5.4.3 Spatial Database Display Tool
PipelineNet's spatial database display tool can
overlay the EPANET model output with various
geospatial properties. The spatial database display
tool of PipelineNet has nineteen criteria to choose
from. The users can select any combination(s) of the
available criteria. For illustration purposes, Figure 7-
21 shows the display of three criteria: oversized pipes
(diameter >30 inches), current monitoring locations,
and low velocity (velocity < 0.001 FPS) pipes.
7.5.5 Case Summary
The case study demonstrates that the PipelineNet
model can be used to perform a variety of practical
analyses to locate monitoring systems. Three
additional tools are available that enhance
PipelineNet's capability in the areas of emergency
response, mitigation, and normal operations. How-
ever, to effectively utilize the PipelineNet model, the
utility must have a calibrated EPS EPANET-based
network model and utility-specific GIS data. At the
7-17
-------
A Reference Guide for Utilities
SetectAHour.rnii Seteciyatocftv\felue:
Tracing Areas (Pipes)
Pressure Areas (Nodes)
Actual Demand (Nodes)
Water Quality (Nodes)
Initial Water Quality
Demand
r Sources of Water
r Sources of Contamination
r Aerial Photo
r Population {8kx* Groups)
f Current Monitoring Staters
Water Quality Cornpainl Areas
" Pipelines
r Schools
Over Sized Pipes Seted Pipe Diameler
Over Sized Pipes
Diameter > 30"
Figure 7-21. Display of Low-Velocity Pipes, Oversized
Pipes, and Current Monitoring Stations Using the Spatial
Database Display Tool.
time this reference guide was being written, the model
is slated to undergo additional enhancements to
improve the following: contaminant database,
consequence assessment, inclusion of time of travel,
conversion from ArcView 3.2 to ArcGIS, and estab-
lishment of links to SCADA data.
7.6 Use of Threat Ensemble
Vulnerability Assessment
(TEVA) Program for Drinking
Water Distribution System
Security
In response to the increased focus on the vulnerability
of drinking water systems to the intentional introduc-
tion of chemical, biological, or radiological contami-
nants, EPA is developing the Threat Ensemble
Vulnerability Assessment (TEVA) Program. TEVA,
when completed, will be capable of analyzing the
vulnerabilities of drinking water distribution systems,
measure public health and economic impacts, and
design and evaluate threat mitigation and response
strategies. TEVA is a probabilistic framework for
assessing the vulnerability of a water utility to a
variety of contamination attacks. Monte Carlo
simulations generate ensembles of scenarios, and
statistics are analyzed to explore the feasibility of
scenarios, identify vulnerable areas of the water
distribution network, and analyze the sensitivity of
the model to various parameters.
The TEVA team includes several individuals from
various organizations. The key EPA TEVA leads are:
Regan Murray, Robert Janke, and Jim Uber.
Key Phrases to Characterize Case Study: hydraulic
and water quality modeling, placement of monitors,
vulnerability assessment, exposure modeling,
contamination assessment, contamination response
tools, probabilistic analysis, economic impact
assessment, threat mitigation strategies.
7.6.1 TEVA Overview
TEVA incorporates a probabilistic framework for
analyzing the vulnerability of drinking water distri-
bution systems. Figure 7-22 outlines the major
components of the framework: the stochastic model-
ing of scenarios, the analysis of potential impacts, and
the assessment of threat mitigation strategies. To-
gether, these three components present an integrated
view of the vulnerability of a unique distribution
system to a wide variety of contamination threats and
the potential for a water utility to decrease this
vulnerability through a set of mitigation strategies.
Preliminary design and implementation has been
completed for a core set of components. A longer-
term research effort is being planned for the other
components.
Stochastic Modeling Scenarios
Select Scenario
41
Simulate Scenario
^
Er
1
semb
e Database
N
Impact Analysis
Public Health
Impacts
Economic Impacts
Identification of
Vulnerable
Populations,
Regions, and
Services
Threat Mitigation Analysis
Evaluation of
Countermeasures
Assessment of Risk
Reduction Strategies
[
Figure 7-22. Threat Ensemble Vulnerability
Assessment Framework.
Without specific intelligence information, one cannot
predict exactly how a terrorist group might sabotage a
water system. Therefore, TEVA is based on a probabi-
listic analysis of a large number of likely threat
scenarios. While the number of possible variations on
7-18
-------
A Reference Guide for Utilities
scenarios is nearly infinite, the vulnerability of the
system can be assessed by selecting a "large enough"
set of likely scenarios. TEVA creates a threat en-
semble, or a set of contamination scenarios, based on
varying the type of contaminant, the amount and
concentration of the contaminant, the location of the
contaminant introduction into the distribution
system, and the duration of the contamination event.
The vulnerability of the system is based on an
assessment of the entire threat ensemble. The
following subsections present an overview of the
aforementioned three key modeling elements.
7.6.1.1 Stochastic Modeling
The stochastic modeling element involves three steps:
selection of the threat ensemble, simulation of the
ensemble, and storage of the output in the ensemble
database. The threat ensemble is a collection of
scenarios that will be simulated. One scenario may
represent, for example, the injection of a 55-gallon
drum containing a biotoxin mixture into one node of
a particular distribution system with 1,000 nodes.
This scenario can be repeated for each of the 1,000
nodes, generating a threat ensemble of 1,000 sce-
narios. One could vary other parameters, such as
contaminant type, quantity, concentration, location,
or duration, to generate other threat ensembles.
Current work is determining how to best select a large
enough threat ensemble in order to accurately assess
vulnerability. While a larger number of scenarios will
allow for the consideration of more threats, con-
straints on computation time require that the number
of scenarios be minimized.
Next, each scenario in the threat ensemble is simu-
lated using a network hydraulic and water quality
model. In TEVA, an EPANET based network model is
generated with all necessary data for running the
simulations. The EPANET model currently includes
first order decay of constituents. Soon to be com-
pleted upgrades to EPANET will allow modeling of
the fate and transport of multiple dissolved constitu-
ents in distribution systems (Uber et al., 2004a).
These upgrades will permit the modeling of reactions
at the pipe wall and in the bulk flow and enable the
inclusion of chemical reaction products, thereby
resulting in more accurate estimates of human
exposure and health risk. The results of the stochastic
modeling of the threat ensemble are stored in the
ensemble database, allowing for later analysis of the
data in the other components of TEVA.
7.6.1.2 Impact Analysis
The Impact Analysis element uses the data stored in
the ensemble database to estimate likely public health
impacts and economic impacts. Public health impacts
include injuries, disease, illness, and deaths. People
can be exposed to contaminants from ingestion of
water, inhalation of volatilized chemicals or particles,
and/or contact with the skin. Depending on the
contaminant, specific dose-response models can be
employed to estimate the various health endpoints.
For many threat agents, reliable data for such models
are lacking, and the ensuing uncertainty in the results
must be understood. For contagious diseases,
dynamic models of disease transmission also must be
included in order to accurately assess health impacts.
Economic impacts include restoration costs (cleanup,
treatment, remediation, and decontamination), denial
of service costs (providing alternative sources of
water), and other costs, such as medical costs (hospi-
talization, vaccines). Psychological costs related to
consumers' loss of trust in the water supply system are
very difficult to estimate. The distribution of impacts
will be calculated from the ensemble database,
thereby providing an estimate of the expected impacts
for the ensemble of threat scenarios.
7.6.1.3 Threat Mitigation Analysis
The Threat Mitigation element explores various
mitigation strategies such as the use of early warning
systems (sensors and data analysis tools), operational
approaches (chlorine boosters, back-up equipment),
and emergency response methods (isolation of part of
the system, public notification). A range of mitiga-
tion strategies can be evaluated with the TEVA
simulations using health risk and economic impact
analyses to rank and select the best alternative for a
set of scenarios (Uber et al., 2004b). This will enable a
quantitative analysis of the benefits of implementing
various strategies.
7.6.2 Application of TEVA to a Water
Distribution System for Optimal
Monitoring
The TEVA computational framework (Murray et al.,
2004, Uber et al., 2004b, Murray et al., 2005) has
been applied to three sizes of distribution systems,
each differing in population by approximately one
order of magnitude. The results shown for this case
study illustrate that TEVA has the potential to help
water utilities assess the contaminants to which they
are most vulnerable, identify the most vulnerable
regions of their distribution systems, and select the
most appropriate mitigation strategies for their system.
Many different forms of contamination monitoring
systems have been proposed, using water quality
sensors, composite or grab sampling, and various
numbers and locations of sensors. Any contamination
monitoring and surveillance program will be budget
constrained. Optimizing the placement of a fixed
number of monitoring stations plays an important role
in the design of the monitoring system. Selecting the
best locations for conducting a routine sampling
7-19
-------
A Reference Guide for Utilities
program to serve as a monitoring and surveillance
program for detection of intentional contamination
can be considered an integer linear programming
problem in which a quantity is optimized subject to a
set of constraints (Berry 2005). Formal optimization
methods and heuristic methods have been applied to
solve such problems. In this case study, a Greedy
heuristic algorithm is used to locate monitoring
stations, given a defined budget or number of
monitoring stations, in order to minimize health
impacts. The following is an example of using TEVA
to evaluate and optimize (in a limited fashion) the
design of a contamination monitoring system for a
water distribution system.
7.6.2.1 Simulation Overview
For the purposes of this analysis, an "all-pipes"
EPANET network model for the sample distribution
system was generated. There are 1,062 miles of pipe
represented in this network model. This system
contains approximately 12,000 nodes, has an average
daily demand of approximately 20 million gallons,
and an estimated population of 130,000. Approxi-
mately 6,000 potential sampling locations were
selected randomly from the nearly 12,000 nodes by
considering that each node had a 50 percent probabil-
ity of inclusion. This pruning was used to make the
problem less computationally intensive and empha-
size that an optimal placement of monitoring stations
will likely be difficult given legal, financial, or
logistical constraints for placing and managing
monitoring stations.
To simulate a contamination scenario, many param-
eters must be specified, including characteristics of
the contaminant, the contaminant-introduction
scenario, and the consumption patterns of the
population. In order to represent the range of possible
parameter values, the TEVA computational framework
uses simulation to vary parameters, such as contami-
nant type, quantity, concentration, location, rate, or
duration, to generate threat ensembles (collections of
many threat scenarios) which collectively can be
analyzed for health impact statistics. All nodes (a
grouping of service connections) in the distribution
system are considered equally likely introduction
points. Biological and chemical contaminants can be
considered, and contaminant introduction can last
from a few minutes to hours to days. For the purposes
of this analysis, a biological agent was considered as
the contaminant and the introduction duration was 24
hours, at a rate of 8.675 liters per hour, and a mass rate
of 1.45 x 10" organisms per minute.
The health impacts are affected by factors such as
dose-response relationships, lethal doses, time-to-
onset of symptoms, time for effective medical
intervention, and the time delay for transporting and
analyzing samples in laboratories. Health impacts to
a population will increase with an increase in the time
required to implement an effective response for a
known contamination event. Considering these
factors, modeling and simulation analyses, such as
those presented here, must be performed on a contami-
nant specific basis. The health impact statistics can
include mean infections/illnesses or mean fatalities.
Infection/illness is a function of the dose of organisms
or contaminant ingested and the probability of illness
caused by that dose, as determined from the contami-
nant dose-response curve (in this case, Salmonella).
Mean infections/illnesses are statistically determined
from the probabilistic analysis of all threat scenarios.
Maximum infections resulted from introduction at the
node delivering the maximum health impacts.
Although there was one worst case node, there were
approximately 60 threat scenarios (contaminant
introduction locations), which delivered at least 50
percent of the maximum lethality. The maximum
number of lethalities was approximately 13,000.
In this TEVA-simulated analysis, the benefits of two
composite grab sampling programs were evaluated
(daily and every 48 hours) and compared to the
benefits provided by a system of real-time (inline,
contaminant specific) sensors. The benefits of the
sampling or monitoring programs are measured by the
hypothetical reduction in public health impacts. In
this analysis, sample location designs are based on
minimizing the mean number of fatalities for 2
sampling frequencies: 24 hours and 48 hours. Six
sampling/sensor station placement scenarios were
evaluated in this analysis: 5, 10, 15, 20, 30, and 40
locations for each program. The Greedy algorithm
used in this analysis will provide an optimal solution
for minimizing public health impacts.
For the purposes of this analysis, the grab samples are
considered to be filtered samples. Filtered samples
represent composited samples that have been col-
lected and concentrated through a filtration device to
better enable the collection and analysis of biological
organisms. Also, real-time water quality monitors are
assumed to detect chemical contaminants or biologi-
cal organisms through the change in water quality,
such as determined by the reduction of chlorine
residual or increase in turbidity. These real-time and
sampling-based analyses are considered ideal,
meaning that detection limits were zero and the
biological contaminant was always detected.
7.6.2.2 TEVA Analysis Approach
This analysis considers attacks at every non-zero
demand node, totaling approximately 10,000. Only
non-zero demand nodes are considered because they
represent service connections that are using water
from the distribution system on a regular basis and,
7-20
-------
A Reference Guide for Utilities
therefore, could possibly be used for contaminant
introduction. Statistically analyzing the approxi-
mately 10,000 threat scenarios provides an estimate of
the hypothetical health impacts in terms of average
health impacts (e.g., fatalities or illnesses) and
maximum impacts. Average impacts could be
expected to result if a saboteur had no knowledge of
where best to attack and simply randomly chose a
node location for contaminant introduction. Maxi-
mum health impacts correspond to a relatively small
set of contaminant introduction node locations (threat
scenarios) that maximize health impacts to the
associated receptors.
The contaminants were modeled as tracers, i.e., free of
hydrolysis, chlorination, pipe wall, or biofilm reac-
tions, which would largely decrease the contaminant's
effectiveness in causing harm to public health (an
extended version of EPANET is undergoing testing to
allow multi-species modeling of contaminants).
Contaminants are modeled using a mass injection
rate, zero volume added, which consequently does not
influence the hydraulic properties of the network, i.e.,
flow increase, decrease, or reversal of flow.
Hydraulic and water quality simulations were run for
192 hours. The disease-causing agent or contaminant
was considered to be a hypothetical, biological
contaminant that is expected to cause infection (50
percent of the time) in an adult when 10,000 or more
of the organisms are ingested. The incubation period
was assumed to be 24 hours, and the time for effective
treatment was 48 hours after the onset of symptoms.
After 72 hours, people either recovered or died. A
sigmoidal dose-response curve was assumed for the
ingestion of organisms with the untreated fatality rate
at 16 percent of those infected.
7.6.2.3 TEVA Analysis Results
Figure 7-23 compares the reduction in mean infec-
tions provided by a routine 24-hour filtered sampling
program, a routine 48-hour filtered sampling program,
and a real-time, continuous, monitoring program.
Similarly, Figure 7-24 compares the reduction in the
maximum number of infections of the same 3 monitor-
ing programs. Again, this scenario assumes contami-
nation by an individual who understands distribution
systems and has the knowledge and resources to
determine the maximum impact location(s).
It is assumed that the computer simulations and the
monitoring/surveillance programs are successful in
reducing public health impacts by preventing further
consumption after detection. The results show that
there is not a significant difference between daily (24-
hour) sampling and 48-hour sampling in terms of
reducing health impacts. As expected, the continuous
monitoring program detects the incident much earlier
48 Hour Sampling Frequency (12 Hour Delay)
- 24 Hour Sampling Frequency (12 Hour Delay)
- Real-Time Monitoring (Hourly Monitoring) 12 Hour Delay I n Response
- Real-Time Monitoring (Hourly Monitoring) 4 Hour Delay in Response
5 10 15 20 30
Number of Sampling/Sensor Stations
40
Figure 7-23. Comparison of 24-Hour, 48-Hour, and Real-
Time, Continuous Contamination Monitoring Systems for
the Reduction in Mean Infections for a 24-Hour
Contaminant Attack,
o
I
umbero
arof Pers
^ A
| i,
re
— • — 48 Hour Sampling Frequency (12 Hour Delay)
— * — 24 Hour Sampling Frequency (12 Hour Delay)
— * — Real-Time Monitoring (Hourly Monitoring) 12 Hour Delay
— x — Real-Time Monitoring (Hourly Monitoring) 4 Hour Delay
80000 -
50000 -
\
\\
\\
\!:=*= =*= =•= -* 5
\ ""* * — ~^— -
--^^
Tf
X X ~ -_,
0 5 10 15 20 30 40
Number of Sampling/Sensor Stations
Figure 7-24. Comparison of 24-Hour, 48-Hour, and Real-
Time, Continuous Contamination Monitoring Systems for
the Reduction in the Maximum Number of Infections for a
24-Hour Contaminant Attack.
than the daily sampling program. Figure 7-23 shows
an 80 percent reduction in mean infections is
achieved using 40 real-time monitors, as compared to
having zero monitors. The results show that it is
important that the real-time program be integrated
with a response protocol. A comparison of the two
real-time, continuous monitoring cases (12-hour delay
versus 4-hour delay in notifying the public) illustrates
the importance of response time. Shortening the time
needed to implement an effective response to reduce
further exposure is critical for reducing the number of
additional infections.
The results of this case study also illustrate that, for
this distribution system, strategically placing just 5 or
10 sampling/sensor stations as part of a monitoring
and surveillance system can have a significant effect
on reducing potential health impacts from intentional
contamination.
7-21
-------
A Reference Guide for Utilities
7.6.3 Case Summary
TEVA is an integrated system intended to provide the
capability for analyzing the vulnerabilities of
drinking water distribution systems and to measure
the public health and economic impacts. TEVA can
be used to design and evaluate threat mitigation and
response strategies related to events of intentional
introduction of chemical, biological, or radiological
contaminants into drinking water networks. Monte-
Carlo simulations generate ensembles of threat
scenarios to identify vulnerable areas of the water
distribution network. TEVA is based on the use of an
all pipes EPANET network model.
7.7 Field Testing of Water-
Distribution Systems in
Support of an Epidemiologic
Study
This case study is focused on the use of information
collected as part of field studies to assist in the
calibration of a hydraulic and water quality model
of a distribution system. The information presented
in this section is based on an ongoing investigation
by the ATSDR at the U.S. Marine Corps Base, Camp
Lejeune, NC (Camp Lejeune). These data are being
collected to support an ongoing epidemiologic
study at Camp Lejeune. The resulting calibrated
model is needed to perform a historical reconstruc-
tion of the water system for the period of interest.
This case study highlights the field methodologies
employed to generate the information proposed for
use in the calibration of the model.
Key Phrases to Characterize Case Study: field
studies, historical reconstruction, hydraulic and water
quality modeling, model calibration.
7.7.1 Case Study Overview
ATSDR is conducting an epidemiologic study to
determine if there is an association between exposure
to contaminated drinking water and birth defects
among children of women who lived at Camp Lejeune
while they were pregnant between 1968 and 1985.
Because of the paucity of historical water distribution
system operational data, information based on the
operation of present-day water distribution systems
will be used for historical reconstruction. Present-day
system operations will be modeled using water-
distribution system models. To calibrate the models
against hydraulic and water quality parameters, field
testing is being performed to gather data and informa-
tion on hydraulic, fate and transport, and operational
characteristics of the water distribution systems
(Maslia et al., 2005; Sautner et al., 2005). The
specific field activities are discussed later in Section
7.7.2.
Figure 7-25. Water Distribution Systems Serving U.S.
Marine Corps Base, Camp Lejeune, NC.
Camp Lejeune encompasses an area of about 164
square miles, and is located in Jacksonville, Onslow
County, North Carolina, bordering the Atlantic Ocean.
The focus of the epidemiologic study is on exposure
from water-distribution systems that historically
served the military base's housing—Camp Johnson,
Tarawa Terrace, Holcomb Boulevard, and Hadnot
Point (see Figure 7-25). Presently, there are two
operating water treatment plants (WTPs) that provide
water for the distribution systems of interest to the
epidemiologic study: (1) the Holcomb Boulevard
WTP that services the Camp Johnson, Tarawa Terrace,
and Holcomb Boulevard areas of the distribution
system, and (2) the Hadnot Point WTP that services
the Hadnot Point area of the distribution system.
Hadnot Point was the original WTP and at one time,
serviced the entire base. The Holcomb Boulevard
WTP presently services the rest of the military
housing areas. A third plant, the Tarawa Terrace WTP,
historically serviced the Tarawa Terrace and Camp
Johnson areas, but this plant was shut down and
replaced by a ground storage tank at Tarawa Terrace
that receives water directly from the Holcomb
Boulevard WTP.
System pressures range from about 55-68 psi
throughout the distribution systems. As topogra-
phy is very flat, ranging from sea level to less than
40 ft, hydraulic heads range from 140-160 ft
resulting in a very mild hydraulic gradient. There
are nine elevated storage tanks in the Holcomb
Boulevard and Hadnot Point WTP service areas.
The range in water level fluctuation for the el-
evated storage tanks is small; generally 1-6 ft.
Excellent mapping information and a detailed GIS
provide good information on the location and
characteristics of the water system facilities.
SCADA data are available that provide continuous
7-22
-------
A Reference Guide for Utilities
Figure 7-26. Continuous Recording Pressure Logger
Mounted on Brass Shutoff Valve and Hydrant
Adapter Cap Used for Fire-Flow and C-Factor Tests.
information on plant discharges and tank water
levels. However, individual buildings and resi-
dences are not metered.
7.7.2 Field Work
A variety of field activities are being performed to
provide a better understanding of the operation of
the water system and to provide information that
can be used to calibrate a detailed water distribu-
tion system model. To date, these activities have
included:
• conducting C-factor and fire-flow tests,
• recording system pressures and storage tank
water levels over time,
• tracer and associated travel time tests, and
• recording continuous flow information at key
locations.
Several of these field activities were performed in
tandem in order to provide an integrated understand-
ing of the system operation and performance.
7.7.2.1 C-Factor and Fire-Flow Tests
C-factor tests and fire-flow tests were conducted in
August 2004 at various sites at Camp Lejeune.
Continuous pressure loggers (Figure 7-26) set to
record pressure at 1-minute intervals were attached to
hydrants. Standard analog pressure gages were also
used as backup during the tests. Hydrant flows were
measured using pitot gages installed on hydrants that
were flowed during the tests. One of the pitot gages
was integrated with a diffuser and cage to both diffuse
the flow from the hydrant and to trap any solids to
prevent damage from the flow (see Figure 4-6 in
Chapter 4). The other pitot gage was a standard gage
attached to the hydrant.
Standard C-factor testing procedures were used to
measure data needed to calculate the Hazen-Williams
C-factors for eight sections representing a variety of
pipe materials and diameters.
Fire-flow tests are frequently used in the process of
calibrating a hydraulic water distribution system
model. One or more hydrants are opened and flowed
to increase flows in the distribution system in the
vicinity of the hydrants. Since friction losses increase
exponentially, the higher flows can result in a
significant lowering of the hydraulic grade line
(HGL). In calibrating the model, the model is applied
under the flow and operational conditions experi-
enced during the fire-flow test and the pressures or
hydraulic grade line observed in the field are com-
pared to the model results. If there are significant
differences between the model and field results,
adjustments are made in model parameters in order to
reduce the differences or calibrate the model. In the
simplest configuration, a single hydrant is flowed and
pressure read at another single hydrant. An alterna-
tive approach was used at Camp Lejeune to improve
the labor efficiency and to collect more data. Con-
tinuous recording pressure gages were installed at up
to six hydrants in the area being tested. Additionally,
pitot gages were installed on two hydrants that were
designated as hydrants to be flowed. Pressure was
measured under several conditions: (1) static condi-
tions at start of testing, (2) while each of the two
hydrants was flowed separately, (3) while both
hydrants were flowed together, and (4) static condi-
tions at the end of testing. The results of such a test at
one site are shown in Figure 7-27 and Table 7-2.
7.7.2.2 Tracer Test and Continuous
Measurements
A field test was conducted May 24-27, 2004, in the
Hadnot Point (Camp Lejeune) distribution system
consisting of three activities: (1) injecting liquid
CaCl2, 35 percent by weight, into the transmission
main on the distribution system side of the WTP to
achieve an elevated conductance and chloride
concentration, and recording conductivity and
chloride concentration using continuous recording
s^
J^fr-WO-Q?
Figure 7-27. Location of Fire Hydrants Used in Fire-Flow
Test at Site H02.
7-23
-------
A Reference Guide for Utilities
Table 7-2. Field Data Collected During Fire-Flow Test at Site H02
Static case
(start)
Hydrant 1
flowed
Hydrant 1+2
flowed
Hydrant 2
flowed
Static case
(end)
FF-H02-PO
Pressure, Psi
53.1
41.4
29.7
43.9
53.5
FF-H02-P1
Pressure, Psi
50.7
37.3
24.5
40.7
51.2
FF-H02-P2
Pressure, Psi
56.2
46.8
36.5
48.1
56.5
FF-H02-P3
Pressure, Psi
52.6
42.9
32.7
44.1
52.9
FF-H02-Q1
Flow, gpm
0
773
631
0
0
FF-H02-Q2
Flow, gpm
0
0
579
747
0
1.0 psi = 6.8948 kPa; 1 gpm = 0.0639 L/S
water-quality monitoring data loggers, (2) injecting
a sodium fluoride solution into the transmission
main to achieve an elevated fluoride concentration
and monitoring fluoride concentration in the
distribution system, and (3) monitoring distribution
system pressures with continuous recording data
loggers attached to selected hydrants and flows and
tank water levels from SCADA data. In addition to
continuously recording tracer concentrations and
conductivity, grab samples were collected for
quality assurance and quality control (QA/QC)
purposes. Samples were analyzed at the Hadnot
Point WTP by ATSDR staff and then also shipped
to the Federal Occupational Health (FOH) environ-
mental laboratory in Chicago, Illinois, for analysis.
Twenty-seven hydrants were selected in the Hadnot
Point area as monitoring locations. For monitoring
conductivity and chloride and fluoride concentra-
tions, nine hydrants were equipped with the Horiba
W-23XD dual probe ion detector (Figure 7-28). For
monitoring conductivity, nine hydrants were
equipped with the Horiba W-21XD single probe ion
detector, thus providing a total of 18 monitoring
locations for continuously recording conductivity
data. For pressure measurements, nine hydrants
were equipped with continuous recording Dixon
PR300 pressure data loggers (Figure 7-26).
Results from the chloride injection were used to
estimate arrival times of the tracer at different
locations throughout the Hadnot Point WTP area.
Of special interest are the extremely long arrival
times—in excess of 16 hours—in the northwestern
part of the of the Hadnot Point WTP area (Figure 7-
29, loggers C01, C02, and F01). Additionally, a
comparison of arrival times of the calcium chloride
tracer at logger location C04 with arrival times at
loggers F04, F05, and F02, led investigators to
consider that there may have been undocumented
closed valves in the distribution system (the closed
valves did not affect C-factor measurements). Post-
test field verification by water utility staff con-
firmed the locations of closed valves, as indicated
by the "•" symbol in Figure 7-29.
7.7.3 Additional Test Procedures
A second tracer test was conducted in the Holcomb
Boulevard WTP area in September-October 2004. In
this test, the normal fluoride feed was turned off for a
period of a week and then turned back on and
Figure 7-28. Horiba W-23XD Dual Probe Ion
Detector Inside Flow Cell.
7-24
-------
A Reference Guide for Utilities
Figure 7-29. Arrival Times of the Calcium Chloride Tracer
at Monitoring Locations in Hadnot Point WTP Area, May
25, 2004.
monitored for another week. Nine locations in the
distribution system were equipped with the Horiba W-
23XD continuous recording, dual probe ion detector
data logger. Minimal labor was required in support of
this test. Additionally, 16 magnetic flow meters have
been installed throughout the system and will be used
in conjunction with future tracer tests to provide
additional calibration information.
7.7.4 Case Study Summary
This case study presents results from preliminary
field-test activities used to gather hydraulic and water
quality data at Camp Lejeune. Field tests to date have
included: (a) recording system pressures and storage
tank water levels, and (b) conducting C-factor, fire-
flow, tracer, and travel time tests. The test data are
being used to assist with hydraulic and water quality
model calibration. They are also being used to plan
and carry out a more refined, detailed field test of
water distribution systems serving military base
housing. These activities will assist in providing
much-needed model parameter data for calibrating
models of the present-day water distribution system.
The present-day models are needed as a first step in
reconstructing historical operations during the period
between 1968 and 1985, as part of an ongoing
epidemiologic study of childhood diseases at Camp
Lejeune.
7.8 Chapter Summary
The case studies presented in this chapter illustrate
the various ways in which the tools presented in
this reference guide can be used. The case studies
also demonstrate that, for each application, a
specific analysis methodology needs to be devel-
oped depending upon the study objectives and
available data.
References
Aral, M. M., J. Guan, M.L. Maslia, J.B. Sautner, R.E.
Gilig, J.J. Reyes, and R.C. Williams, "Optimal
Reconstruction of Historical Water Supply to a
Distribution System: A. Methodology." Journal of
Water and Health, 2(3):123-136, 2004a.
Aral, M. M., J. Guan, M.L. Maslia, J.B. Sautner, R.E.
Gilig, J.J. Reyes, and R.C. Williams. Optimal Recon-
struction of Historical Water Supply to a Distribution
System: B. Applications." Journal of Water and
Health, 2(3): 137-156, 2004b.
Bahadur, R., WB. Samuels, and J. Pickus. Case Study
for a Distribution System Emergency Response Tool.
Published by AWWA Research Foundation. 2003.
Berry, M. Letter (unpublished) to Steve Jones,
ATSDR Region 2 from Michael Berry, Environmental
Health Service, New Jersey Department of Health.
August 31, 1995.
Berry, J., L. Fleischer, W. E. Hart, C. A. Phillips, and J.
P. Watson. "Sensor Placement in Municipal Water
Networks." Journal of Water Planning and Resources
Management. ASCE, vol. 131. 2005.
Clark, R.M., E.E. Geldreich, K.R. Fox, E.W Rice, C.H.
Johnson, J.A. Goodrich, J.A. Banik, F. Abdesaken.
"Tracking a Salmonella serovar typhimurium
outbreak in Gideon, Missouri: Role of Contaminant
Propagation Modeling." Journal of Water Supply
Research and Technology - Aqua. Vol.45, No. 4, pp
171-183. 1996.
EPA. Stage 2 Disinfectants and Disinfection
Byproducts Rule, Initial Distribution System Evalua-
tion. Draft Guidance. 2003.
Fennel, H., D.B. James, and J. Morris. "Pollution of a
Storage Reservoir by Roosting Gulls." Journal of
Society of Water Treatment Exam. 23:5-24. 1974.
7-25
-------
A Reference Guide for Utilities
Geldreich, E.E., K.R. Fox, J.A. Goodrich, E.W. Rice,
R.M. Clark, and D.L. Swerdlow. "Searching for a
Water Supply Connection in the Cabool, Missouri
Disease Outbreak of Escherichia coli 0157:H7."
Water Research 84:49-55. 1992.
Grayman, W.M., R.M. Clark, and R.M. Males.
"Modeling Distribution System Water Quality:
Dynamic Approach." Journal of Water Resources
Planning and Management, ASCE, 114(3). May
1988.
Grayman, W.M., R.M. Clark, B.L. Harding, M.L.
Maslia, and J. Aramini. "Reconstructing Historical
Contamination Events." Water Supply Systems
Security. Editor L. Mays. Published by McGraw-Hill,
NY. pplO.1-10.55. 2004.
Grayman, W.M., A. Ostfeld, and E. Salamons. "Red
Team-Blue Team Exercise for Locating Monitors in
Distribution Systems." Proceedings, ASCE-EWRI
World Water & Environmental Resources Congress,
Anchorage, AK. 2005.
Harr, J. A Civil Action. Published by Random House.
NY. 1995.
Maslia, M. L., J. B. Sautner, andM. M. Aral. "Analy-
sis of the 1998 Water-Distribution System Serving the
Dover Township Area, New Jersey: Field-Data
Collection Activities and Water-Distribution System
Modeling." Agency for Toxic Substances and Disease
Registry, Atlanta, GA. 2000.
Maslia, M. L., J. B. Sautner, M. M. Aral, R. E. Gillig, J.
J. Reyes, and R. C. Williams. "Historical Reconstruc-
tion of the Water-Distribution System Serving the
Dover Township Area, New Jersey: January 1962-
December 1996." Agency for Toxic Substances and
Disease Registry, Atlanta, GA. 2001.
Maslia, M. L., J. B. Sautner, C. Valenzuela, W M.
Grayman, M. M. Aral, and J.W Green, Jr. "Use of
Continuous Recording Water-Quality Monitoring
Equipment for Conducting Water-Distribution System
Tracer Tests: The Good, the Bad, and the Ugly."
Proceedings, ASCE-EWRI World Water & Environ-
mental Resources Congress, Anchorage, AK. 2005.
Murphy, P.J. "Water Distribution in Woburn, Massa-
chusetts." The Environmental Institute, University of
Massachusetts. Amherst, MA. 1986.
Murray, R., R. Janke, and J. Uber. "The Threat
Ensemble Vulnerability Assessment Program for
Drinking Water Distribution System Security."
Proceedings, ASCE-EWRI World Water & Environ-
mental Resources Congress, Salt Lake City, UT. 2004.
Murray, R., J. Uber, and R. Janke. "A Model for
Estimating the Acute Health Impacts Resulting from
Consumption of Contaminated Drinking Water."
Proceedings, ASCE-EWRI World Water & Environ-
mental Resources Congress, Anchorage, AK. 2005.
NJDHSS. "Case-Control Study of Childhood Cancers
in Dover Township (Ocean County), New Jersey."
NJDHSS, Division of Epidemiology, Environmental
and Occupational Health. Trenton, NJ. 2003.
Ostfeld A., and E. Salomons. "A Stochastic Early
Warning Detection System Model for Drinking Water
Distribution Systems Security." Proceedings, ASCE-
EWRI World Water & Environmental Resources
Congress, Salt Lake City, UT. 2004.
Rossman, L.A., R.M. Clark, and WM. Grayman.
"Modeling Chlorine Residuals in Drinking Water
Distribution Systems." Journal of Environmental
Engineering, 120(4):803-820. July/August 1994.
Sautner, J. B., M. L. Maslia, C. Valenzuela, W. M.
Grayman, M. M. Aral, and J. W. Green, Jr. "Field
Testing of Water-Distribution Systems at U.S. Marine
Corps Base, Camp Lejeune, North Carolina, in
Support of an Epidemiologic Study." Proceedings,
ASCE-EWRI World Water & Environmental Re-
sources Congress, Anchorage, AK. 2005.
Uber, J., L. Rossman, and F. Shang. "Extensions to
EPANET for the Fate and Transport of Multiple
Interacting Chemical or Biological Components."
Proceedings, ASCE-EWRI World Water & Environ-
mental Resources Congress, Salt Lake City, UT.
2004a.
Uber, J, R. Janke, R. Murray, and P. Meyer. "A Greedy
Heuristics Model for Locating Water Quality Sensors
in a Water Distribution System." Proceedings, ASCE-
EWRI World Water & Environmental Resources
Congress, Salt Lake City, UT. 2004b.
Walski, T.M., D.V Chase, D.A. Savic, W. Grayman, S.
Beckwith, and E. Koelle. Advanced Water Distribu-
tion Modeling and Management. Haestad Press,
Waterbury, CT. 2003.
7-26
------- |