Office of Research and Development
Homeland Security Research Program
EPA/600/R-17/405 | September 2017
www.epa.gov/homeland-security-research
United States
Environmental Protectior
Agency
Drinking Water Treatment Source Water
Early Warning System State of the Science
Review Report

-------
EPA/600/R-17/405
September 2017
Drinking Water Treatment Source Water
Early Warning System
State of the Science Review
by
Tim Bartrand
Corona Environmental Consulting
Rockland, MA 02370
Walter Grayman
Walter Grayman Consulting
Oakland, CA 94611
Contract Number EP-C-11-037, T-12
Terra Haxton
U.S. Environmental Protection Agency
Water Infrastructure Protection Division
National Homeland Security Research Center
Cincinnati, OH 45268

-------
Disclaimer
The U.S. Environmental Protection Agency (U.S. EPA) through its Office of Research and
Development funded and managed the research described herein under EP-C-11-037, T-12 to
Tetra Tech. It has been subjected to the Agency's review and has been approved for publication.
Note that approval does not signify that the contents necessarily reflect the views of the Agency.
Any mention of trade names, products, or services does not imply an endorsement by the U.S.
Government or U.S. EPA. The U.S. EPA does not endorse any commercial products, services, or
enterprises.
The contractor role did not include establishing Agency policy.
11

-------
Table of Contents
Disclaimer.	ii
List of Tables	v
List of Figures	v
List of Appendix Tables	v
List of Appendix Figures	v
Abbreviations	vi
Acknowledgments.	viii
Executive Summary	ix
1.0 Introduction	1
1.1	Definition and Goals of Early Warning Systems (EWSs)	1
1.2	A Brief History of Source Water EWSs	1
1.3	Types of Incidents and Conditions that Source Water EWSs Address	2
1.4	Components of an EWS	3
2.0 Review of Early Warning System (EWS) Applications	5
2.1	Establishing the Need for EWS	5
2.2	EWS Applications	6
2.3	EWS Case Studies	12
2.3.1	Metropolitan Washington Council of Governments (COG)/Potomac	12
2.3.2	Ohio River Valley Sanitary Commission (ORSANCO)	13
2.3.3	Delaware Valley EWS	15
2.3.4	River Dee EWS	16
2.3.5	West Virginia American Water (WVAW)Utility	18
3.0 Sensor and Monitoring Technologies	20
3.1	Background	20
3.2	Emerging Technologies	22
3.2.1	Routine Online Water Quality Monitoring	23
3.2.2	Biomonitoring	24
3.2.3	Spectral Instruments	26
3.2.4	Cyanobacteria and Cyanotoxins Monitoring	29
3.3	Design and Siting of Monitoring Networks for Source Water EWSs	35
4.0 Statistical Event Detection Methodologies	36
4.1	Background	36
4.2	Integrating Data from Multiple Sensors	36
in

-------
4.3 Event Detection Algorithm Studies	37
5.0 Modeling as an Element of Early Warning Systems	41
5.1	Background	41
5.2	Taxonomy of EWS Models	42
5.2.1	Physically based Models	42
5.2.2	GIS-based Models	43
5.2.3	Data-driven Models	45
5.2.4	Simplified Modeling Techniques	46
6.0 EWS Data Integration and Communication	47
6.1	Data Integration	47
6.2	Communications	48
6.3	Institutions	49
7.0 Conclusions and Research Needs	51
7.1	Conclusions	51
7.2	Research Needs	52
7.2.1	Contaminant Information	52
7.2.2	Monitoring Technologies	53
7.2.3	Placement of Monitors	53
7.2.4	Event Detection Methodologies	54
7.2.5	Fate and Transport Models	54
7.2.6	Data Management and Visualization Tools	55
8.0 References	56
Appendix A — Detailed Case Studies	69
Appendix A Table of Contents
A 1. Danube and Tisza River Basin Case Study	70
A 2. Delaware Valley Case Study	75
A 3. Great Lakes Case Study	80
A 4. Lake Huron to Lake Erie Corridor Case Study	84
A 5. Lower Mississippi River Basin Case Study	88
A 6. Nile River Basin Case Study	92
A 7. Ohio River Basin Case Study	95
A 8. Susquehanna River Basin Case Study	99
A 9. Upper Mississippi River Basin Case Study	103
IV

-------
List of Tables
Table 1. Early Warning System (EWS) Summaries	7
Table 2. Select Approaches for Detecting Chemical and Radiological Threats to Drinking Water
(modified from Brosnan, 1999; Gullick et al., 2003)	21
Table 3. Online Sensor for Cyanobacteria Literature Summary	33
Table 4. Molecular Method Application for Cyanobacteria Detection	34
List of Figures
Figure 1. Potential sources of contamination that could affect a drinking water intake	3
Figure 2. Design features of an integrated early warning system (EWS) (U.S. EPA, 2005)	4
Figure 3. Proposed protocol for development and application of artificial neural network (ANN)
models (from Wu et al., 2014)	39
List of Appendix Tables
Appendix Table 1. Drinking Water Cyanobacteria Alert Levels (adapted from Newcombe et al.,
2010)	82
List of Appendix Figures
Appendix Figure 1. Danube and Tisza River Basins (WWF, 2002)	 70
Appendix Figure 2. Danube River transnational monitoring network (ICPDR, 2015b)	72
Appendix Figure 3. Delaware Valley water basin (Anderson, 2015)	75
Appendix Figure 4. Delaware Valley monitoring stations (adapted from Anderson, 2015)	77
Appendix Figure 5. Western Lake Erie sample locations	81
Appendix Figure 6. Lake Huron to Lake Erie corridor (Morrison, 2006)	 84
Appendix Figure 7. Lake Huron to Lake Erie drinking water monitoring network (Wrubel,
2014)	85
Appendix Figure 8. Lower Mississippi River Basin (Missouri DNR, 2016)	88
Appendix Figure 9. EWOCDS monitoring locations as of 2012 (Louisiana DEQ, 2012)	90
Appendix Figure 10. Nile River Basin (World Bank, 2000)	 93
Appendix Figure 11. ORSANCO ODS monitoring locations (Schulte, 2014)	96
Appendix Figure 12. Susquehanna River Basin EWS (SRBC, 2012)	100
Appendix Figure 13. Upper Mississippi River Basin (U.S. Fish and Wildlife, 2015)	104
v

-------
Abbreviations
AAS
atomic absorption spectroscopy
ANN
artificial neural network
BBN
Bayesian Belief Network
BTEX
benzene, toluene, ethylbenzene and xylene
CBOD
carbonaceous biochemical oxygen demand
CDOM
chromophoric dissolved organic matter
COD
chemical oxygen demand
COG
Council of Governments
CSO
combined sewer overflow
DAEWS
Danube Accident Emergency Warning System
DBP
disinfection byproduct
DDM
Data-driven model
DHS
U.S. Department of Homeland Security
DNA
deoxyribonucleic acid
DO
dissolved oxygen
DOC
dissolved organic carbon
DOM
dissolved organic matter
DRBC
Delaware River Basin Commission
DWMAPS
Drinking Water Mapping Application to Protect Source Waters
EEM
excitation-emission matrix
EWOCDS
early warning organic contaminant detection system
EWS
early warning system
FDOM
fluorescent dissolved organic matter
GA
genetic algorithm
GC
gas chromatograph(y)
GC-FID
gas chromatography flame ionization detector
GC-MS
gas chromatography - mass spectrometry
GIS
geographic information system
GNOME
General NOAA Oil Modeling Environment
HAA
Haloacetic Acid
HAB
harmful algal bloom
HEC-RAS
Hydrologic Engineering Center's River Analysis System
IC
ion chromatography
ICP-MS
inductively coupled plasma - mass spectrometry
ICPDR
International Commission for the Protection of the Danube River
ICP-MS
inductively coupled plasma - mass spectrometry
ICPRB
Interstate Commission of the Potomac River Basin
IC Water
Incident Command Tool for Drinking Water Protection
ILSI
International Life Sciences Institute
LC
liquid chromatography
LC-MS
liquid chromatography - mass spectrometry
LEPC
Local Emergency Planning Committee
LPCF
linear prediction coefficient filter
MCHM
4-methylcyclohexanemethanol

-------
MVE
minimum volume ellipsoid
MWRDGC
Metropolitan Water Reclamation District of Greater Chicago
NASA
National Aeronautics and Space Administration
NATO
North Atlantic Treaty Organization
NHD
National Hydrography Dataset
NO A A
National Oceanic and Atmospheric Administration
NRC
National Response Center
ODS
Organics Detection System
ORP
oxidation-reduction potential
ORSANCO
Ohio River Valley Water Sanitation Commission
PA DEP
Pennsylvania Department of Environmental Protection
PAH
polycyclic aromatic hydrocarbon
PARAFAC
Parallel Factor Analysis
PCB
polychlorinated biphenyl
PCR
polymerase chain reaction
PLSR
partial least squares regression
POC
particulate organic carbon
PWD
Philadelphia Water Department
RAIN
River Alert Information Network
RNA
ribonucleic acid
ROC
receiver operating characteristic
RSI
Risk Science Institute
RTU
remote terminal unit
RWQMN
Remote Water Quality Monitoring Network
SCADA
supervisory control and data acquisition
SMS
Short Message Service
SRBC
Susquehanna River Basin Commission
SRS
surveillance and response system
TEVA
Threat Ensemble Vulnerability Assessment
THM
trihalomethane
TOC
total organic carbon
TSS
total suspended solids
UASI
Urban Area Security Initiative
UMR
Upper Mississippi River
UMRBA
Upper Mississippi River Basin Association
U.S.
United States
USACE
United States Army Corps of Engineers
USCG
United States Coast Guard
U.S. EPA
United States Environmental Protection Agency
USGS
United States Geological Survey
UV
ultraviolet
UV-vis
ultraviolet-visible
voc
volatile organic compound
wssc
Washington Suburban Sanitary Commission
WVAW
West Virginia American Water

-------
Acknowledgments
The U.S. Environmental Protection Agency acknowledges the support provided by the following
individuals in the development of this document:
•	Jonathan Burkhardt, U.S. EPA, Office of Research and Development
•	Jim Goodrich, U.S. EPA, Office of Research and Development
•	John Hall, U.S. EPA, Office of Research and Development
•	Regan Murray, U.S. EPA, Office of Research and Development
In addition, the contributions provided by the source water early warning systems representatives
are acknowledged, including:
•	Kelly Anderson, Philadelphia Water Department
•	Jennifer Heymann, West Virginia American Water
•	Scott Powers, Fairfax Water
•	Lisa Ragain, Metropolitan Washington Council of Governments
•	Jerry Schulte, Ohio River Valley Water Sanitation Commission (ORSANCO)
•	Ian Skilling, United Utilities
•	Ian Warburton, ALS Environmental
The contributions of additional staff at Corona Environmental Consulting under contract EP-C-
11-037, T-12 are also recognized in the development of this document.
Technical review of this document was provided by the following individuals:
•	Steve Allgeier, U.S. EPA, Office of Water
•	Anne Mikelonis, U.S. EPA, Office of Research and Development

-------
Executive Summary
In the United States, customers expect and receive an adequate supply of high quality water
when they turn on their taps. However, under some relatively rare circumstances, contaminants
might find their way into the drinking water resulting in unacceptable water quality. One
important element in the control of water quality is the detection of contaminants in the water
prior to its delivery to the customers. Early warning systems (EWSs) have been developed to
coordinate and systematize these activities. This report is a state-of -the-science review of source
water EWSs. The report evaluated several key studies conducted in the early 2000s to establish
the current state-of-the-science and practice for source water EWSs. The report also identifies
key research areas that need to be addressed to improve EWS.
The first modern EWS was formed after a significant leak of carbon tetrachloride from a
chemical tank into the Kanawha River moved downstream into the Ohio River in 1977. Other
EWSs were established around the world in response to different contamination incidents.
Following the terrorist attacks in New York City and Washington, DC on September 11, 2001,
emphasis in the area of EWSs shifted to concerns over contamination of drinking water
distribution systems and resulted in robust research and development, and implementation of
warning systems in many distribution systems. Almost 40 years after the carbon tetrachloride
spill to the Kanawha River initiated the interest in EWSs, a chemical spill to the Elk River in
West Virginia just upstream of the Kanawha River in 2014 has reinvigorated the interest in
surface water contamination EWSs.
Contamination incidents have been caused by a wide range of sources including industrial and
transportation related spills, non-point sources and urban runoff, intentional contamination and
natural processes. EWSs encompass much more than just sensors or monitors; rather they
include mechanisms for detecting, characterizing, communicating and responding to
contamination incidents in order to initiate effective response actions, and reduce and mitigate
the impacts. The general state-of-the-science of sensor and monitoring technologies, event
detection methodologies, contamination incident modeling tools, and data integration and
communication are presented. EWSs have been implemented around the world as a mechanism
for detecting the presence of contaminants or water quality anomalies in surface waters. The
characteristics and practices of 8 domestic EWSs and 6 international EWSs are summarized in
this report. Detailed descriptions are provided for nine of the most robust systems worldwide.
Although significant research has been conducted on the separate components of an EWS (e.g.,
monitoring technologies, event detection methodologies, modeling tools), additional research
needs to be conducted to evaluate EWSs as a whole to better understand their performance,
detection capabilities and limitations. In addition, future research needs were identified as part of
this study for each of the components and key needs are summarized below.
•	To improve the effectiveness of a source water EWS, more information is needed on the
contaminants that might be a possible threat. This could include developing tools that
enable better access to contaminant information in the watershed.
•	Monitoring technologies research should focus on determining the best parameters to
monitor, understating the field performance of various monitoring technologies,
evaluating monitoring technologies through bench, pilot and field scale testing, and
developing more reliable, practical, and accurate monitoring technologies.
IX

-------
Placement research could help identify where monitors should be located with the source
water to be the most effective for the purposes of the early warning system.
Fate and transport research should focus on models that could be used to support source
water early warning systems. This could include developing approaches to link real-time
data with surface water modeling and simulation tools, incorporating the whole
watershed into the models, and developing linkages between the fate and transport
models and GIS databases.
Detection methodology research needs are associated with the application of EDS to
source water applications. These needs could include better understand current false
positive detection rates and what causes them, developing libraries of events/alarms
associated with common contaminants, and developing additional EDS techniques that
could explore the use of artificial neural networks.
EWS requires data management and visualization tools to support analytics and
communication. Some research needs identified in the study include the development of:
better data transmission tools to support monitor at remote sensing locations, enhanced
data analysis and visualization tools to support real-time response actions, and a reliable
method for validating data from online instruments in real-time.

-------
1.0 Introduction
Customers expect that when they turn on the tap, they will receive an adequate supply of high
quality water. In most situations in the United States, that expectation is met. However, under
some relatively rare circumstances, contaminants might find their way into the drinking water,
resulting in unacceptable water quality.
Generally, the pathways that water follows from source to tap are complex and lengthy. Surface
water or groundwater moves through natural and/or constructed conveyance to a collection point,
where it is delivered to a water treatment plant in which various forms of treatment are applied.
The treated water then enters a distribution system where it is delivered to customers via piping,
pumps, valves and tanks. Within the natural and constructed delivery system, there are many
opportunities for contaminants to enter the water and degrade the quality of the water.
Two important elements in the control of water quality delivered to customers is the detection of
contaminants in the water, and the treatment or other intervention prior to its delivery to the
customers. Detection can be through monitoring or observation. Intervention can be through
increased treatment, through management of the water to keep the contaminated water from
being delivered to the customer, or through issuing "do not drink," "do not use," or "boil water"
warnings until the contaminants have been eliminated or reduced to an acceptable level. Early
warning systems (EWSs) have been developed to coordinate and systematize these activities.
This report is a state-of-the-science review of source water EWSs. The report updates several
key studies conducted in the early 2000s to establish the current state of the science. Monitoring
and contamination warning systems within distribution systems and wastewater systems have
been widely studied and are addressed in this report when they can contribute to the
understanding of source water EWSs.
1.1	Definition and Goals of Early Warning Systems (EWSs)
EWSs have been developed to detect a wide range of natural or human induced incidents
including earthquakes, landslides, tsunami, volcanic eruptions, floods, epidemics, wildfires,
harmful algal blooms (HABs) and contamination incidents. The commonality across the
spectrum of incident types is that early warning systems generate information that empowers
decision makers to take action in time to avoid or mitigate human health risks, economic losses,
or other bad outcomes of disasters and hazardous conditions. If well integrated with risk
assessment studies and with communication and action plans, early warning systems can lead to
substantive benefits (UNEP, 2012). When applied to contamination incidents, EWSs are
intended to identify low-probability/high-impact contamination incidents in sufficient time to be
able to safeguard the public (Storey et al., 2011).
1.2	A Brief History of Source Water EWSs
An incident in the Ohio River Basin in 1977 led to the development of one of the first modern
EWSs to combat source water contamination. A significant leak of carbon tetrachloride, from a
chemical storage facility to the Kanawha River, moved downstream into the Ohio River over a
period of several months. At the time, routine monitoring was not conducted on the Ohio River
that would have detected this chemical. Rather, its presence was discovered when Ohio River
water in Cincinnati, Ohio, was tested for carbon tetrachloride as part of a United States
Environmental Protection Agency (U.S. EPA) research project. The incident led to the
1

-------
establishment of the Ohio River Valley Water Sanitation Commission (ORSANCO) Organics
Detection System (ODS) (Hadeed, 1978).
As was the case with the EWS established on the Ohio River, other EWSs were established
around the world in response to contamination incidents. For example, in 1986, a fire at the
Sandoz Company in Switzerland resulted in a large chemical spill in the Rhine River and the
subsequent implementation of monitors and an EWS on the Rhine River. Other major EWSs
were established in Japan, Canada, the Netherlands and other places around the world. A state of
the art of source water early warning systems at the end of the twentieth century, as documented
in Brosnan (1999), Grayman et al. (2001) and Gullick et al. (2003), is presented in Chapter 2.
Following the terrorist attacks on New York City and Washington, DC on September 11, 2001,
emphasis in the area of EWSs shifted to potential intentional contamination of water distribution
systems. Research, development and implementation of warning systems in distribution systems
was robust. The name of such warning systems was changed to contamination warning systems
in recognition that warnings based on detection of contaminants already in the distribution
system would likely not be early enough to prevent all exposure. The U.S. EPA has recently
referred to such systems as water quality surveillance and response systems (SRS) to reflect the
broader mission of such systems in detecting and responding to water quality threats (U.S. EPA,
2015).
Almost 40 years after the carbon tetrachloride spill to the Kanawha River initiated the interest in
EWSs, a chemical spill to the Elk River in West Virginia, just upstream of the Kanawha River,
has reinvigorated the interest in surface water EWSs (Bahadur and Samuels, 2015).
1.3 Types of Incidents and Conditions that Source Water EWSs Address
Contamination that can affect drinking water sources could originate from many types of
incidents. These incidents could be one-time spills of short duration, an ongoing discharge, or a
recurrent contaminant incident based on seasonal or meteorological/hydrologic conditions. The
following is a list of some of the potential types of contamination incidents:
•	Industrial spills: facility leaks, tank rupture/leakage
•	Transportation related spills: ships, trucks, barges, loading facilities
•	Urban runoff: combined sewer overflows, surface runoff
•	Non-point sources: agricultural runoff, urban runoff, erosion
•	Intentional contamination: terrorists, vandals, illegal disposal of hazardous substances
•	Natural occurrences: algae blooms, organic material (particularly disinfection byproduct
precursor materials)
•	Treatment facilities: insufficient treatment, malfunctioning due to power losses, flooded
facilities
Figure 1 depicts the different types of potential contamination sources as well as the monitoring
locations that could be associated with an EWS.
2

-------
Chemical
Storage
Railroad
Accidents
Automobile
Accidents
Discharges
River Barge
Accidents
Intake
Monitoring
Buoy
Diffuse
Discharges
Intake
Monitor
Drinking
Water
Treatment
Plant
Figure 1. Potential sources of contamination that could affect a drinking water intake.
1.4 Components of an EWS
Grayman et al. (2001) and Gullick et al. (2003) described the following components of an EWS:
•	Detection is a mechanism for recognizing the likely presence of a contaminant in the
source water. Detection might include continuous monitoring, sporadic or periodic
monitoring, public reporting of suspected contamination and self-reporting of
contamination incidents. A relatively new development involves automated event
detection software that uses time series information from monitors and other supporting
information to identify anomalous behavior that might indicate the occurrence of a
contamination incident and notify EWS operators.
•	Characterization is the process of determining what happened during a contaminant
incident. Gullick et al. (2003) outlined a six-step process for characterizing contamination
incidents and synthesizing data and other information into knowledge better suited for
use by staff in response to an incident. The six steps proposed were (i) determine the
specific contaminant(s) involved, (ii) identify the contaminant source, (iii) determine the
temporal and spatial variation in contaminant concentration(s) in the source water, (iv)
assess the dynamic behavior of the contaminant in water (mixing and physicochemical
transformation), (v) predict the movement of the contaminant in water and (vi) determine
the effects on the waterway itself.
•	Response coordination is an institutional framework generally composed of a centralized
unit that coordinates the efforts associated with managing the contamination incident.
•	Communication, in this context, is a means to link and transfer information related to the
contamination incident.
3

-------
• Mitigation is a means of responding to the presence of contamination in the source water
in order to reduce or eliminate its impact on water users.
Figure 2 shows a schematic illustrating data flow and utilization in an integrated EWS. (SCADA
stands for supervisory control and data acquisition.)
Contamination
Monitoring
Sensors, grab samples,
placements models

Secure Pata
Transmission
Direct wire, phone line,
radio, satellite

Pata Acquisition,
Validation. Storage,
and Analysis
SCADA, alarm
management, flow
models, geographic
information systems
software, smart systems

Decision Makers,
Notification, and
Response
Decision software, EPA
Response Toolbox,
confirmation testing,
emergency response,
communication
Figure 2. Design features of an integrated early warning system (EWS) (U.S. EPA, 2005).
4

-------
2.0	Review of Early Warning System (EWS)
Applications
Early warning systems (EWSs) have been implemented around the world as a mechanism for
detecting the presence of contaminants or water quality anomalies in surface waters. In this
chapter, the need for such systems is documented along with a review of the characteristics of
many of the EWSs that have been implemented. One specific type of contaminant incident,
HABs, has emerged in recent years as a significant challenge to the drinking water community.
Early warning activities focused on detecting HABs differ significantly from activities focused
on most other water quality contaminants and are addressed in detail in this chapter.
2.1	Establishing the Need for EWS
In order to identify the extent of source water monitoring and EWSs, Grayman et al. (2001)
conducted a survey of drinking water utilities. A large majority of the utilities were located in the
United States with a smaller number located in Canada and the United Kingdom. Primary
emphasis was placed on surface water sources that were considered to be most vulnerable to
short-term contaminant incidents. Additional details on the survey results can be found in
Gullick (2003). Of the 210 utilities that were contacted, 153 responded to the survey. Treatment
plant sizes varied from 0.15 to 1500 million gallons per day (0.0066 to 65.72 cubic meters per
second). A majority of the 153 utilities that responded to the survey had experienced a significant
contamination incident within the previous five years. Many utilities reported inadequate
warning and response time during these incidents. Utility staff were most concerned about
transportation accidents. The most common contaminants were:
•	Oil and petroleum products
•	Algae and bacteria
•	Particulates
•	Ammonia and volatile organics
•	Pesticides/herbicides/insecticides from industrial spills
•	Agricultural runoff
•	Untreated sewage
•	Seasonal urban runoff
Less than half of the utilities surveyed had an EWS although 90% of them viewed these systems
as important in the future. Only 25% engaged in source water monitoring beyond regulatory
requirements.
Brosnan (1999) reported on the results of a two-day workshop convened by the International Life
Sciences Institute's (ILSI) Risk Science Institute (RSI) with 60 scientists from four countries.
The workshop was to determine the state of the science for EWS to identify strengths and
weaknesses of existing technologies and strategies; to raise awareness of transient hazardous
incidents; and to promote research into prevention, detection, mitigation and treatment. The
workshop determined that the most commonly perceived threats included spills of oil and
industrial products from pipelines, tanks and transportation corridors; insecticides and herbicides
from agricultural runoff; and pathogens from untreated sewage runoff or spills. The most
common EWSs in operation were for chemical and radioactive materials, while monitoring
5

-------
systems for microbial incidents were less common. Intentional threats, such as water system
sabotage or bio-warfare, were not common. Most U.S. utilities monitored some source water
characteristics, but these included a limited number of parameters and were generally conducted
no more than once a week.
2.2 EWS Applications
Source water EWSs in the United States date back to the mid-1970s with the development of the
ORSANCO ODS (Hadeed, 1978). It has served as a model in the development of subsequent
regional systems. Over the past 40 years, the use of EWSs for drinking water sources has
expanded with improvements to previously existing systems, implementation of new systems
and ongoing development of future systems. Since the early development of EWSs, their use has
been enhanced with advancements in monitoring technologies, improved modeling and
communications and the inclusion of additional measured constituents; for example,
biomonitoring and advanced remote sensing enable detection and early warning of toxicity and
HABs. The current application of EWSs encompasses multiple large surface water sources with
a significant national coverage area including the Delaware River Basin, Lake Erie, the Lake
Huron to Lake Erie corridor, the Lower Mississippi River Basin, the Ohio River Basin, the
Susquehanna River Basin, the Upper Mississippi River Basin and others. Internationally, the use
of EWSs has also expanded including development in Africa, China and Europe. Table 1
presents an alphabetical list of source water EWSs identified in the literature search for this
report. The table provides a quick summary of the system along with references to find out more
information. The literature search focused on regional systems and not systems in place at a
single drinking water treatment plant or intake. Detailed descriptions of a few of EWSs are
provided in Appendix A of this report.
6

-------
Table 1. Early Warning System (EWS) Summaries
Location
Summary
References*
Canada
North Saskatchewan
River
EPCOR Utilities in Canada uses two stations at intakes to monitor source water for
chemical dosing decision support. EPCOR is a private water, wastewater and power
supplier whose sole shareholder is the City of Edmonton, Canada.
Gullick et al. (2003)
China
Yellow River
Responsibility for water quality monitoring and protection in China falls under the
purview of the Ministry of Water Resources, the Ministry of Environmental
Protection and the National Environmental Monitoring Center. The seven major
rivers in China are the Yangtze River, Yellow River, Pearl River, Songhua River,
Huaihe River, Haihe River and Liaohe River. China's water quality monitoring
network of source water includes more than 100 stations. A five-year initiative
promoting collaboration between the European Union and China led to the
development of an EWS on a section of the Yellow River. Additional EWSs have
been tested and/or implemented at other locations.
Burchard-Levine et
al. (2012); CNEMC
(2009); European
Commission (2012);
Ministry of
Environmental
Protection (2015);
Zhang et al. (2012)
Danube and Tisza
River Basin
Danube and Tisza
Rivers
Managed by the International Commission for the Protection of the Danube River
(ICPDR), the Danube Accident Emergency Warning System (DAEWS) was
implemented in 1997 in Austria, Bulgaria, Croatia, Czech Republic, Germany,
Hungary, Romania, Slovakia and Slovenia; in 1999, the system was extended to
Ukraine and Moldova and in 2005, expansion included Bosnia-Herzegovina and
Serbia (ICPDR, 2016). An EWS is also being explored for the Tisza River Basin, the
largest tributary to the Danube. The existing DAEWS is primarily a communications
network and consists of a partnership of stakeholders, a periodic water quality
monitoring network, an international communication and alert system and a web and
database portal. The proposed Tisza River EWS includes continuous water quality
monitoring with real-time data transmission (VRIC & EI, 2014). More information
on DAEWS is provided in Appendix A. 1.
ICPDR (2011, 2014,
2015a, 2015b,
2016); IWAC
(2001); VRIC and EI
(2014)
Delaware Valley
Delaware and Schuylkill
Rivers
The Delaware River Basin is comprised of the Schuylkill River and Delaware River
watersheds. Home to approximately 8 million residents, the region spans 13,500
square miles in parts of Delaware, New Jersey, New York and Pennsylvania (DRBC
2013). Led by the Philadelphia Water Department (PWD), the EWS was
implemented in 2004. The EWS consists of 88 monitoring stations, and web- and
phone-based incident reporting. It includes 25 water treatment plants and 24
industrial sites through a partnership between 300 participants from 50 organizations.
Anderson (2015);
DRBC (2013);
Duzinski (2008);
Gullick et al. (2004)
7

-------
Location
Summary
References*

The system includes measurement of dissolved oxygen (DO), turbidity, temperature
and conductivity. The Delaware Valley EWS is described in greater detail in section
2.4 and in Appendix A.2.

Great Lakes
Lake Erie
The Great Lakes HABs program is a collaborative effort between scientists at the
National Oceanic and Atmospheric Administration (NOAA) Great Lakes
Environmental Research Laboratory (GLERL) and the Cooperative Institute for
Limnology and Ecosystems Research. Project goals are to provide a 5-day prediction
of the severity and movement of HABs on Lake Erie. Satellite data, in conjunction
with remote sensing buoys and a comprehensive physical monitoring program, is
used to forecast HABs. Four remote sensing buoys collect data every 15 minutes;
physical collection of samples is done weekly at eight locations. A forecast bulletin is
issued up to every two days during the bloom season and an online HAB tracker is
updated daily with a 5-day forecast. Real-time field measurements, laboratory data,
satellite images and bulletins are publically available online. More information on
this system is provided in Appendix A.3.
Stumpf et al. (2012);
NOAA GLERL
(2015a, b, c)
NOAA GLERL
(2016)
Lake Huron to Lake
Corridor
St. Clair and Detroit
Rivers and St. Clair
Lake
Starting in 2006, the Huron-to-Erie Real-time Drinking Water Protection Network
was developed through a partnership between multiple agencies and participants
including the U.S. EPA and the Michigan Department of Environmental Quality. The
coverage area includes nine monitoring sites located at drinking water treatment
plants. Water quality data are logged every 15-30 minutes; monitoring equipment
measures pH, temperature, DO, conductivity, turbidity, oxidation reduction potential,
chlorophyll, organic carbon, gasoline, diesel fuel, waste oils and other industrial
chemicals. The system includes a database, web portal, and a communication and
data sharing network. More information on the Huron-to-Erie Real-time Drinking
Water Protection Network is provided in Appendix A.4.
Howard (2007);
Lichota and
DeMaria (2009);
NexSens
Technology (2016);
Wrubel (2014)
8

-------
Location
Summary
References*
Lower Mississippi
River Basin
Mississippi River
The early warning organic compound detection system (EWOCDS) was
implemented in 1986 for the southern-most portion of the Lower Mississippi River,
covering Louisiana from Baton Rouge to Plaquemines Parish (Wold, 2015). Water
quality data are collected from seven monitoring stations. The EWOCDS monitors
water draining from more than 40% of the continental United States; that water
serves as drinking water source for 30% of Louisiana's population. Each location
includes a gas chromatograph, with samples collected twice per day at most sites;
two stations have continuous sampling. Monitoring stations measure 28 chemical
contaminants including halogenated organic compounds, chlorinated hydrocarbons
and trihalomethanes. Associated costs are covered by the Louisiana Department of
Environmental Quality. In 2014, the EWOCDS benefitted from a settlement between
Exxon Mobil and Louisiana, from which $250,000 was slated for additions and
upgrades to the EWS. More information on EWOCDS is provided in Appendix A.5.
Louisiana DEQ
(2016, 2014, 2012,
2009); Wold (2015);
Waldon et al. (1998)
Nile River Basin
Nile River
The Nile River Basin EWS was developed in 2008 with funding through North
Atlantic Treaty Organization's Science for Peace Program with coordination from
Egypt's Ministry of Water Resources and Irrigation, National Water Research
Center. The EWS consists of a monitoring network and an internal database portal.
The monitoring network consists of eight sites along the Nile River in Egypt. Real-
time water quality monitoring equipment measures pH, DO, temperature,
conductivity, ammonia and nitrate at 15 minute intervals. Data are accessible through
an internal web portal. More information on the Nile River Basin EWS is provided in
Appendix A. 6.
Khan and Khan
(2008); Khan et al.
(2011)
9

-------
Location
Summary
References*
Ohio River Basin
Ohio River
Ohio River Valley Sanitation Commission's (ORSANCO's) organic detection system
(ODS) was developed in 1977. It currently includes 16 stations at water utilities and
industries along the Ohio River from the confluence of the Allegheny and
Monongahela Rivers in Pennsylvania to the Mississippi River in Illinois. The ODS
monitors water draining from greater than 150,000 square miles (388,498 square
kilometers) and that serves as drinking water source for more than 22,000,000
people. Each station is equipped with a purge and trap gas chromatography system
and tests for the presence of 30 purgeable organic compounds above trigger
thresholds on at least a daily basis. The system also includes reporting of spill
incidents from industries, river users and the National Response Center. ORSANCO
coordinates emergency communications among water utilities and industry users
along the river through an electronic bulletin, Short Message Service (SMS)
messaging, email and a website for online data. The agency also manages the travel
time and water quality modeling during a spill incident. The ORSANCO ODS is
described in greater detail in section 2.4 and in Appendix A.7.
ORSANCO (2016a,
2016b); Schulte
(2014)
Rhine River Basin
Rhine River
The International Commission for the Protection of the Rhine operates nine
international stations plus 20 national monitoring stations in Germany, Holland and
Switzerland. The system uses biomonitors extensively.
Gullick et al. (2003)
River Alert
Information Network
(RAIN)
Allegheny,
Monongahela, Beaver
and Ohio Rivers
The RAIN system is a voluntary cooperative effort of drinking water suppliers in
western Pennsylvania and northern West Virginia. The effort includes water quality
monitoring and data management. In addition, the effort includes data sharing among
the participating utilities, state regulators and the general public. Active monitoring
sites are on the Allegheny, Monongahela and Ohio Rivers. Monitored parameters
include dissolved oxygen, conductivity, pH and temperature. In addition to
maintaining a water quality monitoring and early warning capability, the RAIN
system has a significant public education focus, intended to engage the public in
understanding and protecting drinking water resources. The RAIN system maintains
a publicly accessible website allowing visualization of current water quality data.
River Alert
Information
Network (2016)
10

-------
Location
Summary
References*
Susquehanna River
Basin
Susquehanna River
Led by the Susquehanna River Basin Commission (SRBC), the Susquehanna River
Basin EWS was implemented in 2003 and extended in 2006. The current coverage
area in Pennsylvania and New York includes water suppliers serving approximately
700,000 people. Water quality data are collected with real-time data transmission
from nine monitoring points for pH, temperature and turbidity, while TOC,
conductivity and DO are additionally collected at some locations. Online tools enable
water suppliers to access and analyze data with integrated mapping and a time-of-
travel tool. The coupling of the water quality monitoring network with the SRBC's
communication and data-sharing network enables access to the real-time monitoring
data as well as important water-quality data collected by other agencies. More
information on the Susquehanna River Basin EWS is provided in Appendix A.8.
Gullick et al. (2004);
SourcewaterPA
(2015); SRBC
(2012, 2013a,
2013b, 2015, 2016)
United Kingdom
River Dee
Three water companies, including Hyder Lab and Sciences, partnered with the
government to install and operate three monitoring stations on the River Dee.
Routine grab sampling and analysis are conducted at the monitoring locations as well
as online monitoring. The River Dee EWS is described in greater detail in section
2.4.
Gullick et al. (2003)
Upper Mississippi
River Basin
Mississippi River
The Upper Mississippi River (UMR) Basin spans approximately 189,000 square
miles in parts of Minnesota, Wisconsin, Iowa, Illinois and Missouri and is home to
more than 30 million residents (Swanson, 2012). A pilot monitoring station was
operated from 2003-2007. The UMR EWS was led by U.S. EPA and the Upper
Mississippi River Basin Association. It consisted of six real-time monitoring stations
with measurement of temperature, conductivity, DO, pH, turbidity, nitrate, total
organic carbon (TOC), dissolved organic carbon (DOC) and toxicity (biomonitoring).
More information on the UMR EWS is provided in Appendix A.9.
Allen et al. (2014);
Gullick et al. (2003;
2004); Swanson
(2012); UMRBA
(2016, 2014, 2007)
*References are found at the end of the report.
11

-------
2.3 EWS Case Studies
Interviews were conducted with the staff of five EWSs. The EWSs were chosen based on their
different purposes and configurations and based on the availability and interest of the EWS staff
in participating in the project. The intent of the interviews was to develop a deeper understanding
of the purposes and challenges of EWSs and to identify practical constraints to their operation.
Interviews were conducted using a script, though the interview facilitator asked unscripted
follow-up questions when opportunities for additional data collection were presented. U.S. EPA
initiated each interview by describing the purpose of the interview and U.S. EPA's efforts in
source water event detection and early warning. Results from the five interviews are presented
below.
2.3.1 Metropolitan Washington Council of Governments (COG)ZPotomac
Since 2005, Potomac region water providers have worked with the Metropolitan Washington
Council of Governments (COG) to develop and maintain a regional monitoring capability. The
program was originally funded by an Urban Area Security Initiative (UASI) grant and the
original purposes of the EWS were response to 9/11 and development of regional capacity for
event detection and response. Early in the development of the EWS, basic finished water quality
parameters were monitored by utilities drawing water from the Potomac River and by the
utilities' wholesale customers. One participating utility used the Hachฎ GuardianBlueฎ (Hach,
Loveland, CO) unit for managing data and event detection. Subsequently, source water
monitoring was added at key locations along the Potomac River. At present, the system serves all
utilities drawing water from the Potomac River as far as Brunswick, Maryland. Water quality
monitoring currently in place at nine utilities includes Hach panels (measuring basic water
quality parameters), fish monitors, radiation monitoring and online gas chromatography (GC)
monitoring.
Online instruments are maintained on raw water for plants on the Potomac River and in finished
water for some of the systems. Though the system is a regional system, operation of instruments
and transfer of data is done by participating utilities and a significant challenge is instrument
maintenance and communication of results among stakeholders. The Metropolitan Washington
COG coordinates the monitoring efforts, purchases and facilitates maintenance of instruments
and engages in planning and assessment. Instruments are operated by staff at utilities where
instruments are deployed.
2.3.1.1 EWS Specifics
Monitoring is conducted at the City of Leesburg (Virginia), the Washington Suburban Sanitary
Commission (WSSC), the Washington Aqueduct, Fairfax (Virginia), Brunswick (Maryland),
Rockville (Maryland) and Frederick (Maryland). Each of those utilities also maintains additional
online source water monitoring for process control (as opposed to event detection) and not
connected with the regional monitoring program. Instruments in place online at the monitoring
locations include Hach panels, fish monitors, one radiation monitor and, recently, two Inficon
CMS5000 online GCs. With the exception of the online GCs, instruments (including fish
monitors) collect water quality data at 1-minute intervals. Two portable GCs (Inficon Hapsites)
are available for use in incident response or other purposes.
Data management differs by monitoring location. Communication with most of the Hach panels
is via cellular modems and a commercial remote connection service. One of the utilities has
12

-------
included monitoring data in its supervisory control and data acquisition (SCADA) system to
allow local staff to visualize trends and produce time series plots. At that utility, online monitors
are interfaced with the SCADA system via Modbusฎ protocol, which enables communication
between remote terminal units measuring the online water quality data and SCADA systems.
Prior attempts to establish a regional online communication path/network have been subject to
frequent telecommunication failures and improving communication remains a challenge for the
EWS.
Improved coordination of monitoring and data sharing are stated goals for the EWS. Generally,
monitoring locations/equipment are stand-alone and working independently. The Interstate
Commission on the Potomac River Basin (ICPRB) manages data flow and has developed a
formalized program for spill notifications and data sharing. A single industrial partner provides
data directly to ICPRB in the incident of a spill.
2.3.1.2 Prior Experience and Future Development
To date, online monitors have detected relatively harmless plant incidents such as chemical feed
backflows, though challenge testing indicates that both water quality monitors and the fish
monitors respond to changes in water quality quickly. Utilities have separate protocols for
responding to incidents detected by online monitors. At one of the utilities, incidents are
recorded in an electronic logbook and responses are directed from the utility control center.
The greatest current challenges for the EWS are communication and data management and
analysis. At present, online monitoring data are not managed centrally and are managed
differently by the program's partners. Communication includes management of data within
utilities, between utilities, with the EWS and with external organizations such as ICPRB and
incident response centers. For some of the participating utilities, communication needs include
getting data into a SCADA system for improved access and use by plant staff. Protocols for
sharing data among utilities have been written, but have not been assessed fully or implemented.
Near-future development is planned for both the physical system and the administrative
structure. Utilities are interested in expanding the role of monitoring from detection of
contaminants and incidents to more general collection of water quality data for other purposes
such as operations support. Specific interests include use of fluorescence/spectral instruments for
algae detection, detection of hydrocarbons, and detailed monitoring of organic matter.
2.3.2 Ohio River Valley Sanitary Commission (ORSANCO)
ORSANCO is a regional organization that supports utilities in the Ohio River Valley.
ORSANCO manages a regional organics detection system (ODS) as part of their core function.
The ORSANCO's ODS program entails continuous water quality monitoring and contaminant
early warning at a regional scale. The impetus for the ORSANCO ODS was a series of high-
impact contamination incidents on the Ohio River and its tributaries. One of the most important
of those incidents was a 1977 release of carbon tetrachloride in the Kanawha River that impacted
drinking water supplies on the Kanawha and Ohio Rivers including for Huntington, West
Virginia, Portsmouth, Ohio, and Cincinnati, Ohio.
Monitoring equipment used in ORSANCO's ODS is a series of gas chromatographs (GCs)
owned by ORSANCO and operated by participating utilities. As currently configured, the ODS
can be considered a screening program because of limitations in the number of monitoring
locations and the range of parameters that can be monitored routinely. Quality assurance and
13

-------
quality control (QA/QC) procedures were established specifically for the ODS to ensure data
quality without causing excessive demands for utility laboratory staff.
ORSANCO's primary stakeholders are participating utilities. ORSANCO and utilities maintain
open communication and ORSANCO facilitates the transfer of data between utilities.
ORSANCO's overall funding ($2.6-$3 million annually) is from the member states and U.S.
EPA, while the ODS program is funded exclusively by the states. Stakeholders beyond
participating utilities include states and two industrial dischargers (in Parkersburg, West Virginia
and Saint Albans, West Virginia). ORSANCO serves many groups, including the U.S. Coast
Guard (USCG) and other entities involved in spill response. Utilities provide significant in-kind
support as staff time to conduct analyses and facilities to house equipment.
2.3.2.1	EWS Specifics
Organics monitoring is conducted at 16 locations, with 13 on the Ohio River main stem and the
rest on the major tributaries (Kanawha River, Allegheny River and Monongahela River).
Instruments are housed and operated at participating utilities and include gas chromatography -
mass spectrometry (GC-MS) and GC with flame ionization detector (GC-FID) as well as online
GC analyzers. In routine monitoring, samples are analyzed for 30 organic compounds four times
per day and online GC analyzers operate at two-hour intervals. Utilities commit to report all
detections greater than 2 ppb, but generally report detections greater than 1 ppb. Samples are
collected in duplicate for confirmatory or more detailed analysis. During emergency response,
utilities might be asked to analyze additional samples at intervals as short as one hour.
Data are reported to and maintained by ORSANCO. At least every week, the data are
downloaded and reviewed by ORSANCO staff. The reviews are more frequent after/during spills
and following spurious detections. After a spill, data are shared with all participating utilities and
chromatographs are shared with downstream utilities. Chromatographs could also be shared with
state regulatory agencies, though historically states have not requested them. ORSANCO does
not plan to share data with the general public, though states or utilities could choose to share
data. ORSANCO is currently developing a web-based tool for internal data maintenance and
utilization.
In the event of a detection during routine sampling, ORSANCO performs additional QA/QC on
the data. Once results are verified, ORSANCO requests the utility partner to collect an additional
sample and downstream utilities, state regulatory agencies and the NRC are notified. After spills
or other incidents, data might be used in concert with modeling to predict and track contaminant
plumes. The Ohio River main stem and major tributary hydraulic/hydrologic information
(predicted river depths and flows for the next five days) are provided to ORSANCO by the U.S.
Army Corps of Engineers (USACE) on a daily basis (weekdays only except during flood
conditions when data might also be provided on weekends). These data are generated by the
USACE CASCADE model (currently being transitioned to the HEC-RAS model) and used by
ORSANCO as input to predictive river water quality models when a spill occurs.
ORSANCO conducts numerous activities related to the ODS and EWS such as bi-monthly
nutrients and metals monitoring, fish and macroinvertebrate monitoring, and combined sewer
overflow (CSO) long-term control plan tracking. Online monitoring for routine water quality
parameters and for cyanotoxins are in development.
2.3.2.2	Prior Experience and Future Development
14

-------
The ODS and ORSANCO are operational and an integral component of Ohio River Valley
emergency response. The ODS has been used in the following incident responses over the past
five years:
•	A 10,000-gallon diesel spill in Cincinnati, Ohio, that was not detected by the ODS.
•	The Freedom Industries 4-methylcyclohexanemethanol (MCHM) spill (Elk River), in
which the ODS detected MCHM at the nearest downstream monitoring location (St.
Albans, West Virginia).
•	A methylene chloride release in Cincinnati, Ohio, that was detected by the ODS. In
addition, the ODS was used to help identify the source.
•	An ethanol spill from derailed train cars, in which the associated diesel fuel was detected
by the ODS, but ethanol was not.
ORSANCO expects the ODS to continue operation for the foreseeable future. ORSANCO
identified a number of improvements and enhancements it is considering. Algae monitoring for
HAB early warning could be added and could include monitoring on reservoirs and the
deployment of multi-parameter probes. ORSANCO is also interested in improved access to
information on contaminants in the Ohio River Valley such as shipping cargo information,
inventories of compounds carried on rail cars, and an inventory of contaminants stored in the
Ohio River Valley watershed.
2.3.3 Delaware Valley EWS
The Delaware Valley EWS was established in the 2004 and 2005 timeframe using grants from
the U.S. EPA and the Pennsylvania Department of Environmental Protection (PA DEP).
Subsequent funding was from a grant from the Maritime Exchange. The EWS has been in
continuous operation since its inauguration and has undergone multiple upgrades. The EWS
monitors source waters in the Delaware River and Schuylkill River basins — the total watershed
area for the source waters for the Philadelphia Water Department (PWD) is approximately
10,000 square miles. At present, the Delaware Valley EWS has more than 325 system users from
Pennsylvania, Delaware and New Jersey. The EWS is maintained through contributions from 13
Pennsylvania water suppliers, four New Jersey water suppliers and 14 industrial water
users/dischargers. The system is owned and operated by PWD, though other users can influence
the operation and development of the system. The Delaware River Basin Commission (DRBC)
collects fees on behalf of PWD for system maintenance purposes. The primary goal of the EWS
is to support existing notification protocols in place to protect the drinking water supply for more
than three million people. The type of notification (email or phone call) and the notification
recipients depend upon the perceived severity of the incident. Additional benefits of the EWS are
that it provides a secure forum for data and information sharing.
2.3.3.1 EWS Specifics
Components of the Delaware Valley EWS include 88 USGS gauge stations linked to the system,
four remote terminal units (RTUs) connected to the water quality monitors and analytical tools.
Monitoring is conducted on the Neshaminy Creek (at an Aqua Pennsylvania drinking water
plant), the Schuylkill River and the Delaware River. Temperature, pH, flow, DO and
conductivity are monitored at each location and data are collected at 15-minute intervals. The
system previously included a fish monitor, but it was removed since it was difficult to maintain.
Additional water quality data available for the Delaware River include monitoring data, from
PWD and American Water treatment plants, and USGS water quality data.
15

-------
Analytical tools in the EWS include a time-of-travel model and a tidal model. The time of travel
model draws data from USGS gauge stations and has been used in spill tracking. The tidal modal
is a critical component of the system because the zone of tidal influence of the Delaware River
extends above drinking water treatment plant intakes.
During emergency response, the EWS provides data to system partners who make decisions
regarding their own operations. At present, the EWS does not include event detection.
Emergency response roles of the EWS are to facilitate communication for entities spanning a
large geographic area and to provide a redundant path for communication.
Data are managed through a web page accessible by system users. When a user enters an
incident into the system, a "code red" is issued and users are notified. Users determine if the
incident/spill that they are reporting is either a low or high-level incident based on their
judgement. Both low and high-level incidents generate an email to system users. High-level
incidents also generate telephone calls. The system does not conduct downstream notification,
though this feature has been requested. Paying members, regulatory agencies, Local Emergency
Planning Committees (LEPCs) and the USCG have access to the Delaware EWS website.
2.3.3.2 Prior Experience and Future Development
The Delaware Valley EWS has detected or reported more than 500 incidents. Those include:
•	A coal fly ash spill in an upper part of the watershed of 100 million gallons in 2007
•	Numerous transportation accidents
•	An industrial fire with runoff potential
•	A crude oil spill of approximately 275,000 gallons in 2004
•	A cyanide chloride compound discharged after wastewater treatment
Following the cyanide chloride discharge, drinking water treatment plant intakes along the flow
path were closed and the EWS was used to facilitate communication during the response.
The Delaware Valley EWS is expected to continue operation for the foreseeable future and is
considering expansion. Water suppliers operating downstream of the current coverage area (e.g.,
on the Brandywine Creek and Christina River in Delaware) are interested in joining. The system
currently has no plans to include public reporting of suspected incidents or the capability for the
public to download and analyze data. This decision was taken to ensure data remain secure,
confidential and accurate. At present, data sharing requests are submitted by email and
considered.
Research gaps and concerns identified by the Delaware EWS include the need for data analysis
tools and for developing stronger linkages between water supplier source water protection efforts
and local emergency planning. The data analysis tools would be used for event detection and for
improved use of water quality data. A significant concern about data analysis tools is the
likelihood of a high frequency of false positive assessments.
2.3.4 River Dee EWS
The River Dee EWS is overseen by United Utilities, a private company that serves as water
supplier and wastewater manager and that has about seven million customers. In general, the
EWS is intended to facilitate pollution risk management on the River Dee., and was formed in
response to a large spill of chlorinated phenols that occurred in the mid-1980s. The spill
impacted about two million people in north Wales and northwest England. Contaminated water
16

-------
entered the drinking water treatment plants and was detected in finished water. An inquiry of the
incident recommended improved coordination among agencies to better manage the risk.
At its inception in 1984, the EWS was administered as a joint operation between the water
companies and government organizations. Water companies contributed both staff and facilities.
The EWS has been continuously operated since 1985. Laboratory analyses, sampling and online
monitoring are conducted by a contractor. Originally four water companies participated in the
EWS, but two of the original companies merged, and, thus, three companies are currently taking
water from the River Dee. The Welsh regulators have the lead for the system. Two regulating
agencies contributed financially at the onset of the effort, and currently provide in-kind support.
The system is maintained financially by payment based on the amount of water taken from the
river by each utility.
2.3.4.1 EWS Specifics
The River Dee EW S includes water quality monitoring (online and grab sampling), centralized
data management and coordination and communication with water utilities and regulators.
Monitoring includes three online water quality stations operating 24 hours per day, grab
sampling at eight locations along the river system and laboratory analysis of grab samples within
6 hour or less. Online monitoring locations (Manley Hall, Poulton and Huntington) were chosen
based upon the location along the river system and the proximity to treatment plants and the
EWS laboratory. Results of laboratory analyses are reported to a quality control officer and
regulators (Natural Resources Wales). When a pollution incident is declared, alarm notices are
circulated among regulators and water companies. EWS laboratories have chemists/analysts who
are engaged in analyzing follow-on sampling.
Online monitoring includes standard water quality parameters (e.g., DO, conductivity, pH, and
temperature); two online monitoring locations employ online volatile organic compound (VOC)
monitors (purge and trap GC). Six of the eight grab sample locations are along the River Dee
main stem, while the remaining two are located on important tributaries. Grab samples are
collected and analyzed twice per day. Routine sampling and analysis for phenols and other target
compounds is conducted at one of the monitoring locations. The EWS employed fish monitoring
in the early days of operation. However, the fish were generally stressed in all water quality
conditions and, thus, the fish could not distinguish between pollution incidents and background
stressors. Other biomonitoring technology was also piloted and abandoned because of
performance or operational problems.
All online monitoring stations are linked via a commercial software product. The software can
generate alarms based on set points, trends and interruption in signal/communication. Alarms can
be based on the excursion of a single observation outside control limits. In routine online
monitoring and reporting, water quality data are recorded each 15 minutes and individual data
are retained for 48 hours. Data older than 48 hours are aggregated in daily minimum, maximum
and median values, and then individual data are discarded. Each 12 months daily median,
minimum and maximum values are summarized in a report to the drinking water company and
other stakeholders.
17

-------
2.3.4.2 Prior Experience and Future Development
The River Dee system meets the regulator goals of addressing water quality problems in
catchments rather than via addition of expensive treatment, which further supports the continued
operation of the River Dee EWS.
Early operation of the EWS resulted in numerous alarms, hundreds per month. These alarms
were reduced significantly by filtering the samples prior to the GC analysis. Per the operators,
about eight alarms occur per month from the intake protection system presently. The main
detected incidents are from ammonia (tracked to sewage plant discharges to the river), DO
swings (often diurnal) and nitrate alarms. At present, roughly half of the monthly alarms are
genuine, and are followed up by human investigation. For example, an ammonia alarm
investigation might involve determination of concentrations of caffeine, cholesterol or other
indicators of sewage. Because the system has been operating for decades and with consistent
staffing, analysts have become experienced and adept at investigating alarms and assessing
whether they are genuine.
Future considerations for the River Dee EWS include adding optical DO monitors and other
optical sensors. Interest in optical sensors is partly driven by the inaccuracy of ion selective
electrodes and operational problems with colorimetric methods. Research gaps identified by the
River Dee EWS include practical and accurate detection of inorganics.
2.3.5 West Virginia American Water (WVAW) Utility
In response to the 2014 Freedom Industries MCHM spill on the Elk River, the West Virginia
legislature requires public water systems, providing water to 100,000 customers or more, to
monitor source waters for key classes of contaminants. The West Virginia American Water
(WVAW) utility opted to implement source water monitoring at all of its eight surface water
plants in West Virginia, even though only one plant serves more than 100,000 people. The rule
requiring monitoring is not specific regarding the classes of contaminants that should be
monitored or the details of monitoring (e.g., instruments, frequency, performance objectives)
required. In response to the monitoring requirement and to augment public health protection and
incident response, the WVAW utility designed, fabricated and tested monitoring panels; installed
panels at each of its treatment plants; established remote communication to the panels; connected
data to an information system (with event detection capabilities); and established procedures for
maintaining instruments and monitoring data. The panels included the online monitoring
instruments. Stakeholders of the system include WVAW utility and regulators. The utility funds
the entire cost of the system.
2.3.5.1 EWS Specifics
The WVAW EWS includes online monitoring, centralized data management, event detection for
water quality changes and instrument performance notifications. Monitoring instruments have
been operational for roughly one year. Monitoring is conducted on raw water from each of eight
water treatment plants. Online monitoring parameters include dissolved organic carbon (DOC),
pH, conductivity, oxidation-reduction potential (ORP), DO, temperature and turbidity. The
precise location of monitors is different for each plant, since the locations were each chosen
based primarily on logistical considerations, such as electricity and communication. For some
plants, water is sampled directly from the raw water pumps. Travel time from the monitoring
location to the treatment plant ranges from minutes to hours, depending on the plant. Grab
18

-------
sampling data are also collected for each plant raw water. Other data used for evaluation of water
quality data and event detection include:
•	Streamflow and precipitation data are accessed from external sources
•	ORSANCO notifications of spills and incidents
•	Additional water quality data from the River Alert Information Network (RAIN) system
(WVAW utility is a member utility)
•	Customer calls
•	Gas chromatography - mass spectrometry (GC-MS) analyses conducted at two treatment
plants
Water quality monitors are connected to data loggers (at each site) and access to data on the data
loggers is via cell modem or Ethernet cable, depending on the availability of service at the
location. Water quality data are maintained via cloud computing and are accessed by WVAW
staff via a commercial web-based data management and analysis tool. At present, data can be
accessed by the WVAW source water protection manager, water quality managers, plant
operators and key staff. Data are analyzed by the Detector event detection software tool
(http://www.mindset-tools.com/?page=detector) (Decision Makers Ltd., Boynton Beach, FL).
Event detection has been underway for less than six months and a full review of results is not yet
available. Detection of incidents by Detector software or by other analyses would result in
confirmatory laboratory analysis followed by appropriate communication within WVAW, with
regulators and with the general public (if merited).
2.3.5.2 Prior Experience and Future Development
The WVAW EWS has been operational for roughly one year and is still under development. A
significant challenge to the system is determining which parameters need to be monitored so as
to target the contaminants of greatest concern at each of the individual treatment plants. Other
sensors might be added based on specific challenges at each plant. For example, an online algae
monitor has been deployed at a plant that has its intake is on the Ohio River. Additional
monitoring locations upstream of intakes are also under consideration, though practical
challenges such as access to communications and power and vandalism must be overcome. One
specific developmental goal is to achieve more consistent system operation. Consistent operation
will require refining instrument operations and maintenance protocols and will require
addressing vulnerabilities in the data communication pathway.
An additional goal is improving data analysis and interpretation. At present, strong connection
between the water quality parameters that can be monitored and the contaminants present in the
water near the drinking water intakes has not been established. Because WVAW utility is in the
early stages of implementation of event detection, experience is required for better interpretation
of alarms and for distinguishing false alarms from consequential water quality changes.
19

-------
3.0	Sensor and Monitoring Technologies
Sensors and monitors are key components of an EWS as a mechanism for detecting the presence
of potential contaminants. In this chapter, the general state of the science of sensor and
monitoring technologies is presented. The design of monitoring networks for use in an EWS
including the selection of technology and the siting of monitors are also discussed.
3.1	Background
Gullick et al. (2003) expanded on a review conducted by an ILSI working group (Brosnan, 1999)
to develop a table of monitoring technologies for use in a source water EWS. The resulting
technologies available when Gullick et al. (2003) conducted their study are presented in Table 2.
The table shows instruments in three cost ranges (low-, medium- and high-cost) for measurement
of classes or groups of threats (contaminants). Negative and positive aspects of each technology
are outlined.
In Table 2, selectivity is noted as a negative for some technologies because the authors assessed
that technologies detecting a broad range of contaminants were preferable to more selective
technologies because any contaminant could be present in source water. An alternative viewpoint
is that, for a particular source water intake, identifying contaminants that are more likely to
threaten the water supply and selectively detecting contaminant levels associated with harmful
levels of exposure, could be appropriate choices for an EWS. The connection between sensor
choice and exposure was also made by Brosnan (1999), who noted that treatment-plant managers
considered the top threats to their water supplies to be pollutants from oil and petrochemical
spills, agricultural runoff, and untreated sewage. Many of the technologies listed in Table 2 are
laboratory instruments used for detection of specific contaminants. Those instruments include
inductively coupled plasma mass spectrometry (ICP-MS) for specific and sensitive detection of
metals, liquid chromatography and liquid chromatography-mass spectrometry (LC and LC-MS)
for specific detection of polar organic compounds, GC (including purge and trap GC) and gas
chromatography-mass spectrometry (GC and GC-MS) for detection of volatile organics, ion
chromatography (IC) for detection of ionic contaminants, and atomic absorption spectroscopy
(AAS) for detection of metals.
20

-------
Table 2. Select Approaches for Detecting Chemical and Radiological Threats to Drinking Water (modified from Brosnan,
1999; Gullick et al., 2003)
Threats
llii>h-cos( inst rii men Is (S100.000s)
Medium-cost instruments ($10,000s)
Low-cosl instruments (SI000s)
Technologies
Pros
C oils
Technologies
Pros
Cons
Technologies
Pros
C oils
Ions (salts)



IC
Fast, broad,
selective

Ion prolvs
Sensili\ e
Select i\ e
Metals
ICI'-MS
last. broad
II).
scnsiti\c
Stall". lah
AAS
Fast, sensitive
Staff, lab
Ion prolvs
Sensili\ e
Select i\ e
Polarography
Fast, fairly
selective
Selective



Polar organics
I.C-MS
Hroad II)
Stall". lah
I.C
Broad ID
Staff, lab
I V

l.ack ol"
scnsili\ il\
T< )C
Broad ID
Lack of
sensitivity



Non-polar
organics
(iC-MS
Hroad II)
Sial'l". lah
I.C
Broad ID
Staff, lab



Volatiles, oils,
hydrocarbons
(i( -MS
Kroad II)
Sial'l". lah
\>&'T-GC
Broad ID
Staff, lab
Smell hell
lasl
1 luman
deleclors
(iC
Broad ID
Staff, lab
I'luorescence
(oil. HC)
Broad ID
Interferences
Specific
compounds
(i( -\IS. I.C-
MS
Kroad II)
Sial'l". lah



lniniunoassa>
ipeslicidesl
I'asl.
specific
Stall'
Biotoxics



Uiomonitors
Continuous,
fast
Lack of
specific ID



Radiation



Tn I mm
Fast, specific
Not available
online



(ianima
detector
Fast, broad
ID, available
online
Lack of
specific ID
Kcta or alpha
detector
Fast
Lack of
specific ID, lah.
evaporation
step, not
available on Ink-
AAS, atomic absorption spectrometry (furnace or flame); Biomonitors, fish, daphnids, mussels, algal fluorescence, and luminescent bacteria, I .road II), can
monitor for many compounds simultaneously; Fast, not quantified in Gullick et al., 2003; GC, gas chromatography; HC, hydrocarbon; IC, ion chromatography;
ICP-MS, inductively coupled plasma mass spectroscopy; ID, identification; LC, liquid chromatography; MS, mass spectrometry; P&T, purge and trap; Selective,
monitors for a single compound; Smell bell, trained staff detect unusual odors in water sample; TOC, total organic carbon; UV, ultraviolet
21

-------
Few technologies suitable for continuous, online monitoring are included in Table 2. As
demonstrated in the response to the Elk River spill of MCHM, highly specific laboratory
analyses can be an important part of a spill response (Rosen et al., 2014), though their expense
and lack of mobility limit their use in routine, high-frequency monitoring.
Gullick et al. (2003) identified the following research and development needs focused on
technology improvements:
•	Continuous monitors for low levels of dissolved oil and petroleum products
•	Rapid automated sensors for established and emerging pathogen and bio-warfare agents
•	Simultaneous identification of multiple pathogens (combined biosensors)
•	Improved sensor sensitivity
•	Continuous, online and remote sensing monitors for a greater number of chemical
parameters
•	Electronic nose improvements
•	Improved biological monitors
•	Technology exchange between water supply and sensor development industries
As noted in the following section, significant progress has been made in addressing these needs
since 2003, particularly in the identification and development of biosensors and the development
of techniques and tools for improved source water assessment. In contrast, significant research
needs remain with respect to more timely detection of microorganisms, acceptance of new or
unfamiliar monitoring technologies, and development of smart sensors.
3.2 Emerging Technologies
Published studies report the emergence of numerous and diverse monitoring technologies in
approximately the last decade. Since the reviews conducted in the early 2000s, two general types
of online monitoring advances have dominated:
•	Development and application of new technologies for detection of constituents of interest
•	Modification of existing technologies to overcome features that limited their ability to be
field-deployed
Many of the studies published over the last decade report performance of novel technologies in
laboratories or other settings that might not reflect conditions representative of those in drinking
source water. A partial list of realities of sensor deployment in source water includes fouling,
interference by matrix constituents, power failures, accessibility difficulties, and degradation of
critical sensor components. Many sensor developers have overcome these problems, indicating
that they are tractable. However, until deployment problems are identified and addressed by both
the sensor vendors and customers, emerging technologies will not be effective components of
EWSs.
Van den Broeke et al. (2014) have developed a web-based compendium of online monitoring
technologies and case studies illustrating their use in water settings (drinking water and
wastewater). Among other uses, the compendium is intended to facilitate matching online sensor
selection to specific operations and applications. The Online Water Quality Sensors and
Monitors Compendium can be found at www.wqsmc.org. Data that can be retrieved from the
compendium for a technology and a particular application include summaries, advantages,
22

-------
disadvantages, acquisition and operational costs, installation, operational and maintenance
information, benefits, manufacturers and suppliers and use cases. Engineers have many
considerations when selecting technology as a component of an EWS and those considerations
extend well beyond the parameter(s) the technology measures.
This section provides an overview of emerging water quality monitoring technologies, with a
focus on online monitoring technologies developed over the last decade. Online refers to
technologies that automatically collect and communicate data, and includes technologies that
monitor a flowing sample as well as those that collect and analyze discrete samples. The
technologies reported in the literature vary widely in their state of development (from conceptual
to commercialization), their focus on contaminants of relevance to drinking source water, and
their potential for field deployment. A review of the literature indicates that the emerging
monitoring technologies most relevant to drinking source water EWSs are biomonitors (monitors
using the response of biological organisms to water constituents) and spectroscopic instruments
(instruments sensing absorbance, transmittance, scattering/reflectance of electromagnetic
radiation). These emerging technologies are the focus of this review. Readers are referred to
recent reviews published by Banna et al. (2014) and O'Halloran et al. (2009) for additional
information on other emerging technologies. Specific mention of sensor or vendor names in this
section should not be construed as an endorsement or criticism of the technology or the vendor.
3.2.1 Routine Online Water Quality Monitoring
As noted by Storey et al. (2011), robust, commercially available technologies exist for many
parameters routinely monitored in source water and treated water, with the notable exceptions of
ammonia and fluoride. Methods for incorporating routine water quality data into EWSs are
described in section 4.2. Wider application of commercially available instruments is governed by
(i) their costs and benefits and (ii) whether these technologies can be used for specific
contaminant detection.
Two recent reports (Hall and Szabo, 2010; Hall et al., 2009) described findings from the U.S.
EPA sensor technology evaluations. Although the U.S. EPA studies focused on monitoring and
detection in treated water, their findings are relevant to source water monitoring because many of
the instruments evaluated could be used in both source and treated water and because the
evaluations included operability and other technology features related to their practical use in a
drinking water treatment environment. Evaluations indicated that free chlorine and total organic
carbon (TOC) were the water quality parameters most sensitive to contaminant presence in
distribution systems in which free chlorine was the secondary disinfectant. TOC is likely an
important parameter for detecting contaminants in source water, too, although natural TOC
variability in source water is much higher than that in treated water.
Several notable advances in monitoring of routine parameters have occurred since the early
2000s. For example, researchers have demonstrated the use of ultraviolet-visible (UV-vis)
spectroscopy for monitoring suspended solids rather than turbidity (Louren<;o et al., 2006; Reiger
et al., 2004). Both studies note that turbidity, though familiar in the water treatment context, is a
surrogate for suspended solids and subject to bias and interference. Another advantage to the use
of UV-vis spectroscopy for determining suspended solids is the potential for replacing several
probes (a turbidimeter and other water quality monitoring devices) with a single probe.
Banna et al. (2014) included differential pH sensors (one probe measures in a buffer and another
in sample) and amperometric sensors among currently-available pH sensors tested by the U.S.
23

-------
EPA. Emerging technologies for pH measurement include use of volume, optical and electrical
changes in hydrogels, potentiometric pH sensors, ion-selective field-effect transistor pH sensors,
and fiber-optic based pH sensors. Potential advantages of the alternative technologies include
greater sensitivity, ability to miniaturize sensors, and greater longevity.
Miniaturization could facilitate easier deployment of sensors, particularly as components of
multi-parameter probes (Gunatilaka et al., 2007). Miniaturization also facilitates sensor
deployment in tighter or more difficult-to-access spaces such as building plumbing systems.
3.2.2 Biomonitoring
A comprehensive review of the many commercially available options for biomonitoring is found
in a recent study by Kokkali and van Delft (2014). Biomonitoring refers to the use of living
organisms that serve as indicators of water toxicity. Here, toxicity is specific to the organism(s)
used in the monitor and does not refer to human toxicity. A wide diversity of organisms is used
in commercially available biological monitors, or biomonitors. Kokkali and van Delft (2014)
separated the organisms into the following broad categories:
•	Microorganisms
•	Enzyme-based detection and mammalian cells
•	Invertebrates
•	Fish
•	Other organisms
•	Multiple species
Commercially available versions of biomonitors include both laboratory and field-deployed
monitors. All field-deployed biomonitors face two major challenges beyond those faced by other
types of water quality monitors:
•	Maintaining a population of viable organisms
•	Mapping the behavior/response of biological organisms to water quality changes relevant
in the drinking water production context
Storey et al. (2011) identified five commercially available biomonitors applicable to online
source water monitoring. The organisms used in these biomonitors included bacteria, algae and
fish. Limitations of commercially available biomonitors differ by technology and include slow
response times, interference of constituents like chlorine with the organisms used for
biomonitoring (i.e., false positives), low sensitivity of organisms to target analytes, lack of
specificity (i.e., response to a broad range of contaminants rather than a targeted substance or
group of contaminants), and operational challenges associated with maintaining the organisms.
Ren and Wang (2010) illustrated the challenges related to choice and maintenance of biological
organisms in their study comparing biomonitors using the planktonic crustacean Daphnia magna
and Japanese madaka or rice fish (Oryzias latipes). The organisms differed in their sensitivity to
contaminants, the clarity with which their responses to stimulation could be measured and the
duration of their survival without food. The authors found that neither species performed well for
all metrics and suggested monitoring with both. Maradona et al. (2012) used responses of four
species - Daphnia magna andHyalella azteca (crustaceans), Lumbriculus variegatus (a
freshwater worm) and Pseudokirchneriella subcapitata (a freshwater algae) - and principle
component analysis of their responses to develop a library of contaminant-specific responses.
24

-------
Tests with atrazine and tributyltin indicated their approach was promising and capable of
detecting the target contaminants within two to four hours, which is likely sufficiently fast in the
context of an EWS. Timescales relevant to spills on river systems are typically on the order of
hours and determined by travel times from spill locations to drinking water plant intakes (though
the travel time was much shorter in the 2014 Elk River spill of MCHM).
As noted in a review conducted by Girotti et al. (2008), many studies have evaluated
bioluminescent bacteria as components of biomonitors. Bioluminescent bacteria offer advantages
over other organisms including the potential for genetic modification, ease in measuring light
output (the means for assessing response), rapid response to exposure to toxic compounds and
response to a wide range of contaminants relevant to drinking water. Girotti et al. (2008) found
studies on monitoring of cyanotoxins, arsenic, toluene, heavy metals, pesticides and polycyclic
aromatic hydrocarbons (PAHs) in surface water. At least one biomonitor, Microtoxฎ CTM, has
been developed to work as an online instrument (http://www.modernwater-
monitoring.com/product-microtox-ctm.html).
Although biomonitoring is often used for general water monitoring, some systems have been
configured for detecting specific contaminants. For example, Zhang et al. (2012) developed an
online biomonitor for the detection of carbamate pesticides. Their biomonitor employed medaka
(iOryzias latipes). Dose-response experiments with 0. latipes revealed a stepwise response to
increasing doses of carbamate pesticides. The study did not report attempts to challenge the
biomonitor with other toxicants and it is unclear how specific the monitor is for detection of
carbamate pesticides.
Recent reviews of whole-cell biomonitoring (Eltzov et al., 2009) and monitoring with
bioluminescent organisms (Woutersen et al., 2011) suggested that those two technologies have a
high potential for being incorporated into future incident detection systems. These two types of
biomonitoring systems can be configured as contaminant-specific systems and have the potential
for online deployment (i.e., in field settings). Whole cells (or other biologically based materials)
used as biosensors can be suspended, entrapped or bound to substrate. Based on a literature
survey, Eltzov et al. (2009) reported application of whole-cell biosensors in a wide variety of
water environments and for a wide variety of contaminants. Whole cell biosensors recognize
(detect) contaminants via bio-recognition elements on cells that have been immobilized on the
biosensing device.
In their review of biosensors based on bioluminescence, Woutersen et al. (2011) suggested that
biosensors with genetically modified luminescent bacteria have the potential to provide real-time
toxicity monitoring in water. Bioluminescent biosensors can provide rapid results that are easily
measurable. Bioluminescence biosensors can signal the presence of a toxicant by "lights out"
response (reduction in luminescence after exposure) or "lights on" response (increase in
luminescence over background level) after exposure to a target contaminant. The lights-off
versions typically use naturally occurring organisms and detect toxicity rather than the presence
of a specific contaminant. The lights-on versions frequently rely on genetically modified
organisms and they can detect a single contaminant or a group of related contaminants.
Woutersen et al. (2011) also reported the use of bioluminescence biosensors for many
contaminants relevant to drinking source water early warning. Those contaminants included
benzene, toluene, ethylbenzene xylene (BTEX) compounds, polychlorinated biphenyls (PCBs),
phenols and metals (in particular, lead, mercury, iron and cadmium). Use of these biosensors as
25

-------
operational components of an EWS would require rigorous determination of their sensitivities,
interferences and operational requirements.
3.2.3 Spectral Instruments
Many studies published in the past decade report novel applications of spectral instruments for
monitoring water quality. This section provides a snapshot of the state of science in the use of
spectral instruments for water quality monitoring by highlighting studies demonstrating online
monitoring or with a focus on detection of parameters of greatest relevance to drinking water
treatment and early warning.
Several commercially available sensors that utilize absorbance at wavelengths in the UV-vis
range have been deployed for continuous water quality monitoring. Etheridge et al. (2013) noted
that portable UV-vis spectrometers have been used for monitoring nitrate nitrogen (NO3-N),
DOC and total suspended solids (TSS). Examples of other studies reporting use of online or
field-deployed UV-vis spectroscopy include monitoring nitrate+nitrite in wastewater (Drolc and
Vrtovsek, 2010), ozone and assimilable organic carbon in partially treated drinking water (van
den Broeke et al., 2008), and DOC and particulate organic carbon (POC) in surface water (Jeong
etal., 2012).
As with other sensors, a significant hurdle that spectral (spectroscopic) instruments must
overcome is operation in the field environment. Etheridge et al. (2013) noted that fouling via
biological growth and chemical precipitation are significant problems that must be overcome for
field deployment of any spectral sensor. In their study, the authors noted significant fouling and
instrument performance degradation over periods of time as short as two weeks. In response, an
antifouling system (limiting exposure of the probe to stream water in periods between
measurements and automated rinses of lenses with clean water) was designed and implemented.
The antifouling system enabled the use of the probe for extended periods with only minor drift in
DOC measurements. Commercially available, field-deployable UV-vis spectrophotometers are
sometimes equipped with fouling control. For example, the s::can spectro::lyser™ spectrometer
can be equipped with automated air cleaning or brushes for physical cleaning (http://www.s-
can.at/en/). Similarly, the Zaps LiquID™ Station spectrophotometer
(http://www.zapstechnologies.com/the-liquid-station/) conducts periodic automatic cleaning of
optical surfaces via pressurized air and clean water.
Analysis of data from UV-vis spectral instruments can be more complex than analysis of data
from other instruments. Some instruments capture absorbance at many wavelengths over a wide
wavelength range. Generally, absorbances must be corrected for suspended solids (which cause
an apparent change in absorbance; Hu et al., 2016) and fouling, and data might be analyzed as
corrected absorbances, first derivatives of corrected absorbances or second derivatives of
corrected absorbances. Finally, constituents other than a target analyte can influence the
absorbance spectrum. Partial least squares regression (PLSR) is the most common technique for
matching spectral signals to target analytes (Chen et al., 2014; Korshin et al., 1997; Langergraber
et al., 2003; Reiger et al., 2004; van den Broeke et al., 2008), though at least one study indicated
that useful information can be drawn from direct use of raw spectral data (Vaillant et al., 2002).
Prior to PLSR, absorbance data could be transformed to enhance the signal for constituents of
concern. For example, Roccaro et al. (2015) found that log-transformation of absorbance data
prior to analysis generated improved correlation of spectral data with trihalomethanes (THMs)
and haloacetic acids (HAAs).
26

-------
Some UV-vis spectral instalments are programmed with laboratory-generated profiles based on
samples in water matrices that could differ from the matrix of the water being monitored.
Because PLSR and other data reduction techniques are not widely used among environmental
engineers, some users likely rely on factory calibration and profiles for their instruments. It is
possible that the lack of calibration in the water being tested could result in suboptimal
performance of an instrument.
In their report on the use of fluorescence spectroscopy for characterizing dissolved organic
matter (DOM) in drinking source water, Carstea et al. (2014) contended that advantages of
fluorescence spectroscopy over other water quality instruments were its high sensitivity, small
required sample volume, low-to-no sample preparation requirements, and short measuring time.
Disadvantages included difficulties in managing and analyzing the complex data stream some
instruments produce and included interference via matrix constituents. Known interferences with
fluorescence spectroscopy include inner filtering effect (absorbance of some emitted energy
within the sample), oxidants and fluorescence quenching due to temperature, pH, and metal ions
(Henderson et al., 2009).
Carstea et al. (2014) reviewed studies of water quality characterization via fluorescence
spectroscopy and listed research studies in which researchers attempted to detect and
characterize the water quality parameters including:
•	Biochemical oxygen demand
•	TOC
•	Nitrogen and chemical oxygen demand
•	DOM
•	Diesel pollution
•	Viral pathogens
•	Pesticides
•	Biological water quality
Fluorescence spectroscopy instruments can be designed to operate at one or several wavelengths
or to produce a three-dimensional matrix of excitation-emission data (the excitation-emission
matrix [EEM]) (Sanchez et al., 2014). Like absorbance spectra, EEMs require sophisticated
analyses for interpretation, given their complexity and the fact that multiple substances can
produce similar signals in the EEM. The most commonly reported technique for analyzing the
EEM is parallel factor analysis (PARAFAC) (Guo et al., 2010; Johnstone et al., 2009; Sanchez,
et al., 2014; Yang et al., 2015). Parallel factor analysis identifies the most important excitation
and emission wavelengths, which can be used to indicate specific contaminants or to indicate
water quality in lieu of analysis of the entire EEM.
The majority of recent studies reporting the use of fluorescence spectroscopy in drinking water
applications were for characterizing and quantifying organic matter. The nature of organic matter
in drinking source water is a key determinant of coagulation efficacy and the potential for
disinfection byproduct (DBP) formation. Sanchez et al. (2014) developed analytical techniques
for evaluating the 3-D EEM and for characterizing DOM changes along a treatment train in a
drinking water treatment plant. In addition, the authors developed techniques to explore the
removal of DOM in coagulation. Over a three-year period, the researchers collected pre- and
post-coagulation samples, and characterized the DOM changes associated with coagulation.
Although the study entailed grab sampling (rather than online monitoring), the study provided a
27

-------
proof-of-concept for continuous monitoring and analysis of the EEM. The study also
demonstrated that fluorescence spectroscopy provided information for enhancing coagulation
that is not provided by simple measurement of TOC/DOC.
Stedmon et al. (2011) determined that online fluorescence spectroscopy could be developed for
early warning of sewage contamination of wells. Sewage was spiked into water samples
corresponding to various levels of treatment and PARAFAC, a multi-way spectra decomposition
method, was used to determine factors associated with the presence of sewage. A single
excitation wavelength and two emission wavelengths appear sufficient for detecting sewage in
that system. Reduction of the EEM to a smaller set of excitation and emission wavelengths
allows development of a practical sensor for in situ monitoring and early warning of well
contamination.
An alternative to using the entire EEM for water quality monitoring is monitoring of
fluorescence at a single wavelength. According to Downing et al. (2012), instruments that
measure chromophoric dissolved organic matter (CDOM) fluorescence at wavelengths of
approximately 460 nm in response to excitation at approximately 370 nm have proven to be a
highly sensitive and useful tool for elucidating spatial and temporal DOM variability. Sensors
measuring the fraction of CDOM that fluoresces (i.e., fluorescent dissolved organic matter
[FDOM]) are commercially available, simpler to use than analyzers using the full EEM, and
increasingly used more in research settings. Downing et al. (2012) assessed the performance of
four commercially available FDOM sensors in laboratory and field studies. The four instruments
differed in performance, but all were susceptible to interference from color, suspended solids and
temperature. These findings indicate that data from the instruments must be corrected for light
scattering and temperature for accurate measurement of FDOM. Those corrections might require
simultaneous deployment of turbidimeters (or other suspended solids monitors) and temperature
probes.
Bridgeman et al. (2015) assessed the feasibility of fluorescence excitation at two sets of
wavelengths for continuous detection of TOC and microorganisms. Assessments were conducted
in a laboratory setting and using river water samples. Bridgeman et al. hypothesized that:
•	Fluorescence emitted at 400-480 nm under excitation at 300-360 nm (fulvic-like
fluorescence) is indicative of the presence of organic carbon (peak C)
•	Fluorescence emitted at 340-370 nm under excitation at 220-240 nm or 270-280 nm
(tryptophan-like fluorescence) is indicative of microbial activity (peak T)
At low TOC concentration (less than 25 mg/L), a linear relationship was observed between peak
C fluorescence and TOC. At higher TOC concentrations, the relationship became nonlinear, but
remained monotonic, though characterized by significant scatter in the data. Correlation between
measures of biological water quality (heterotrophic plate count, flow cytometer counts, counts of
individual species of bacteria) and peak T were not as good as correlation of TOC with peak C.
The experimental design might have been responsible for the ambiguous results for
microorganisms; counts of microorganisms in the samples were not precisely established or
controlled in experiments. In summary, Bridgeman et al. (2015) demonstrated the potential for
fluorescence field instruments to monitor organic carbon, though it is critical to account for
interferences and understand the relationship between peak C and TOC at relatively high TOC
concentrations.
28

-------
By examining the full excitation-emission spectrum, Zhou et al. (2016) determined that
excitation at 275 nm and emission at 342 nm correlated well with concentration of tryptophan-
like proteinous fraction of CDOM in a surface water. The tryptophan-like fraction can be
indicative of sewage in the surface water. The authors hypothesized that variations in CDOM
from sewage inputs to a lake could be distinguished from variations due to hydrologic processes.
Distinguishing an incident (e.g., a sewage overflow) from other natural variations (e.g., those
driven by rainfall and runoff) is a significant challenge in source water monitoring for early
warning. Zhou et al. (2016) used PARAFAC analysis to identify the fluorescence signature
providing the best indication of sewage inputs to a lake and to identify point source pollution
inputs to the lake.
Li et al. (2016) examined the full EEM and determined that excitation at a single wavelength -
280 nm - could excite humic-like and protein-like emission. The humic-like fraction was
subsequently found to correlate well with both THM and HAA yields. This finding is similar to
that of Johnstone et al. (2009) who used PARAFAC to deduce EEM factors that correlated well
with DBP formation potential and DBP production. Li et al.'s (2016) good correlation with DBP
formation potential using a single wavelength for excitation wavelength and a single emission
wavelength indicates potential for development of a simplified online DBP precursor sensor.
3.2.4 Cyanobacteria and Cyanotoxins Monitoring
A specific type of contaminant incident that has emerged in recent years as a significant
challenge to the drinking water community is the presence of HABs. Recent incidents impacting
the operation of treatment plants, such as the City of Toledo operation, and forecasts that climate
change and anthropogenic nutrient loading could lead to more frequent and widespread HABs
with the potential for greater production of cyanotoxins emphasize the significance of this
challenge (Gehringer and Wannicke, 2014; Paerl and Huisman, 2009). The literature review
uncovered numerous reports of early warning activities focused on HABs, highlighting their
importance. For this reason, a more thorough review of these activities is described here.
HABs are caused by the growth of cyanobacteria, which are sometimes referred to as blue-green
algae, although they are bacteria, and not algae. Some cyanobacteria produce cyanotoxins
depending on environmental factors such as nutrient limitation. Several cyanotoxins have
varying health impacts including the nervous system, liver and skin toxicity. The most studied
toxins are microcystin, cylindrospermopsin, anatoxin-a and saxitoxin. Microcystin is the most
frequently detected of the cyanotoxins in U.S. lakes and reservoirs (Graham et al., 2009, 2010;
Loftin, 2008).
An increase in cyanotoxin-producing algal blooms in the United States has been linked with two
major factors. The first is the eutrophication of freshwater sources, caused by nutrients, mainly
nitrogen and phosphorus (Dolman et al., 2012; Yuan et al., 2014). The other is the impact of
climate change, which is producing a warming trend in the majority of lakes (O'Reilly et al.,
2015). Warming trends could lead to increases in HABs, and significant implications for the
monitoring and management of bloom incidents (Delpla et al., 2009; Paerl and Paul, 2012).
Drinking water treatment plants can remove cyanobacteria, and treat cyanotoxins to some extent.
The limited data available show that there is a very low percentage of treated drinking water with
cyanotoxin detections (Carmichael, 2000). The treated water detections that are documented
have made an impact on utilities and regulators. In August 2014, the City of Toledo, Ohio issued
a Do Not Drink/Do Not Boil order that stemmed from a microcystin detection in the treated
29

-------
water above 1 [j,g/L for microcystin, which at the time, was the threshold established by Ohio
EPA (Ohio Environmental Protection Agency, 2014). In the summer of 2015, the U.S. EPA
issued Health Advisories for microcystins and cylindrospermopsin largely in response to the
Toledo incident.
The combination of increasing blooms (both size and frequency) and current attention to
cyanotoxins in drinking water emphasizes the need for early warning and rapid detection
methods for cyanobacteria. Historic monitoring for cyanobacteria and algae has been conducted
by collection of grab samples and enumeration and identification with microscopes. This time
consuming and laborious process is completely unsuited for the kind of rapid detection that is
needed. Three categories of early warning and rapid detection are discussed in the following
sections: remote sensing, modeling and monitoring.
3.2.4.1 Remote Sensing
Remote sensing has been proposed and developed as a component of HAB early warning.
Remote sensing offers advantages over water quality sampling including collection of data over
wide areas and with high frequency relative to traditional water quality sample collection and
analysis. Thus, remote sensing could be used in conjunction with traditional water quality
analysis and other data collection and analysis for early detection of HAB occurrence and
improved drinking water operation response.
Seven satellites with instrumentation suitable for chlorophyll-a, or cyanobacteria detection, are in
orbit now (Trescott, 2012). Remote sensing involves determining absorption and/or reflection at
wavelengths specific to a particular water quality constituent. Chlorophyll-a has distinct
absorbance peaks (433 nm and 686 nm) and reflectance peaks (550 nm and 690-700 nm) in the
visible light portion of the spectrum (Cracknell et al., 2001).
Remote sensing poses challenges to widespread use for cyanobacteria early warning:
•	For lakes and reservoirs, water quality data alone are not sufficient for early warning to
manage HAB risks to drinking water plants, since risks to drinking water intakes depend
on unique characteristics of water bodies, particularly their size, and the practicality of
water quality monitoring with a limited number of sensors.
•	Cyanobacteria blooms can occur on rivers (in addition to lakes and reservoirs), which are
flowing water bodies and are subject to obscuring by tree canopies.
•	Models and analyses connecting water quality and other data used in early warning need
to be developed and validated to account for the unique characteristics of the water body
that influence the absorbance or reflection of the light.
•	Cloud cover can obscure the satellite images.
•	Due to the flyover schedule, the satellites which are used for remote sensing are not
always present above a given location. Satellites can have a fly over schedule as little as
1 to 2 days, or as many as 16 days (Trescott, 2012).
•	The detection of a bloom incident is possible at high biovolume (mass of
microorganisms) levels but is not accurate at the lower levels required for managing risks
associated with drinking water sources.
•	The growth of cyanobacteria does not necessarily mean that toxins are being produced,
and the remote sensing technology cannot determine if cyanotoxins are present (Freeman,
2011).
30

-------
The National Oceanic and Atmospheric Administration (NOAA) has two active HAB prediction
demonstration projects (NOAA, 2013). In the Gulf of Maine, NOAA has been forecasting red
tide. In the Great Lakes, the Great Lakes program has a fully developed prediction network
(Freeman, 2011; NOAA GLERL, 2016), which focuses its efforts on Lake Erie, Saginaw Bay
and Lake Huron.
Several other projects were short-term in nature and did not have fully developed EWSs, or were
centered on oceanic HAB prediction (Hunter et al., 2008; Klemas, 2012; Kudela et al., 2015;
Kutser et al., 2006; Lunetta et al., 2015; Matthews et al., 2010; Simis et al., 2005; Trescott,
2012). NOAA has three additional research projects using remote sensing for HAB forecasting in
coastal locations (NOAA, 2013).
3.2.4.2	Models for Predicting HAB Blooms and Transport
As described in section 5.2, models (analytical or computer simulations) can be a useful
component of an EWS by forecasting or integrating water quality data. Inputs to models can be
data that indicate the extent of blooms (e.g., spectrally-resolved satellite imagery), grab sample
results (e.g., chlorophyll a concentration, cyanobacteria concentration or cyanotoxins identity
and concentration), hydrologic and atmospheric data (e.g., wind speed and water temperature),
and other water quality data related to bloom occurrence (e.g., pH and DO). Lake Erie is the
focus of two physical models that have been developed for predicting the occurrence, fate and
transport of algae blooms (Francy et al., 2015; Wynne et al., 2013). Wynne et al. (2013)
developed their model for algal bloom early detection for Lake Erie. Their model used satellite
imagery as a key input and determined that blooms were predicted by water temperature and
wind speed. Francy et al. (2015) monitored for concentrations of cyanobacteria by molecular
methods, for algal pigments such as chlorophyll and phycocyanin by using optical sensors and
for a number of other water quality parameters that served as inputs to models for various sites
on Lake Erie and some inland lakes. These models demonstrate the potential for successful
development of EWS for drinking source water for incidents that are of growing concern. They
also demonstrate that such systems (and their underlying models) can be data intensive and
require specialized expertise for their successful use.
3.2.4.3	Monitoring Equipment
Online fluorometer analyzers can be used to detect cyanobacteria by measuring the fluorescence
of the phycocyanin pigments (the most specific indicator of cyanobacteria) or chlorophyll-a (an
indicator of algae). The online analyzers are capable of detecting cyanobacterial concentrations
as low as 150 cells/mL. Reported uses of online fluorometers for cyanobacteria detection are
summarized in Table 3. Zamyadi et al. (2012) conducted a partial survey of studies on excitation
and emission wavelengths used in studies of cyanobacteria and phytoplankton monitoring. The
authors reported a wide range in excitation wavelengths (430-625 nm) and a much narrower
range in emission wavelengths (655-690 nm). A principle concern in the study conducted by
Zamyadi et al. (2012) was the utility of fluorescence probes for in vivo monitoring of
phycocyanin and chlorophyll-a, and the use of the fluorescence probes within an EWS. The
authors determined that online fluorescence spectroscopy improved early warning capabilities
over monitoring of other physicochemical properties alone and improved the ability to
distinguish cyanobacterial blooms from other algae blooms. The fluorescence spectroscopic
measurements in the study did not correlate with microcystin concentration.
31

-------
Polymerase chain reaction (PCR) can be used to detect cyanobacterial deoxyribonucleic acid
(DNA) that has the potential to produce toxins. Reviews of the use of molecular methods in
detection and management of cyanobacteria have been published by Moreira et al. (2014) and
Srivastava et al. (2013). Studies reporting the use of PCR for cyanobacteria detection or
detection of toxic cyanobacteria are presented in Table 4. For example, the microcystin
synthetase gene cluster is an indicator of the potential for microcystin production (Francy et al.,
2015). A complication related to use of PCR for detection of toxin-producing cyanobacteria is
that several genera of cyanobacteria can have this gene cluster, but it can also be absent in those
genera. Identification by microscopy can identify the cyanobacteria genera, but cannot
determine if the cyanobacteria is a toxic strain. The ability to differentiate a toxic bloom from a
non-toxic bloom is one advantage of molecular methods over microscopy and online analyzers.
At present, commercially-available PCR-based monitors capable of continuous monitoring are
not available.
32

-------
Table 3. Online Sensor for Cyanobacteria Literature Summary
Location
Finding
References
Ruhr Basin,
Germany
An online fluorometer is used for chlorophyll-a
detection.
Bode and Nusch
(1999)
Murray and
Lower Darling
Rivers, Australia
The project examined whether the in situ
quantification of phycocyanin by fluorometry could
be used to determine the abundance of
cyanobacteria present. Abundance was measured in
the laboratory as biovolume from samples collected
at the same time as the phycocyanin measurements.
The study found a strong positive relationship
between the two measurements. However, it was
found that the use of in situ phycocyanin
fluorometry was not effective in turbid water higher
than 50 Nephelometric turbidity units (NTU) as this
produced false-positive readings for phycocyanin.
Bowling et al.
(2013)
Maine et Loire,
France
Four online analyzers were tested. Although the
results were not as reliable as laboratory tests the
information was very useful for making quick
adjustments to plant operations.
Cagnard et al.
(2006)
Taiwan
Results showed that chlorophyll-;/, turbidity and the
colonial status of the cyanobacteria significantly
interfered with the measurement of phycocyanin
fluorescence. Models were developed to
compensate for the effect of chlorophyll-;/,
turbidity and colony size on the measurement. The
models were successfully used to correct
phycocyanin probe data collected from several
reservoirs in Taiwan to establish good correlation
between measurements made using the
phycocyanin probe and microscopic cell counts.
Chang et al. (2012)
Poland
Statistically significant correlation between
cyanobacterial biovolume and fluorometer readings
and very strong correlation between chlorophyll-a
and fluorometer readings was found.
Izydorczyk et al.
(2009, 2005)
Hong Kong
Online fluorescence monitoring is used in
conjunction with other parameters to send alarms
when a bloom is occurring along the coast so that
additional physical samples can be collected.
Lee et al. (2005)
Canada
A significant relationship between phycocyanin
fluorescence and cyanobacterial biovolume was
found when the growth was dominated by
microcystin producing cyanobacteria.
McQuaid et al.
(2011)
North and South
Carolina
A multipie-fixed-vvavelength spectral fluorometer
was used to measure chlorophyll-a, and correlated
with laboratory measurements. The analyzer
Richardson et al.
(2010)
33

-------
Location
Finding
References

appears to be a useful tool for early warning of
harmful algal blooms.

Canada
In vivo probes were used to trace the increase in
floating cells over the clarifier, a robust sign of
malfunction of the coagulation-sedimentation
process. Pre-emptive treatment adjustments, based
on in vivo probe monitoring, resulted in successful
removal of cyanobacterial cells. The field results on
validation of the probes with cyanobacterial bloom
samples showed that the probe responses are highly
linear and can be used to trigger alerts to take
action.
Zamyadi et al.
(2014)
References are listed at the end of the report.

Table 4. Molecular Method Application for Cyanobacteria Detection
Location
Finding
References
Mai pas Dam,
New England
region of
Australia
Showed that bloom components can be identified
and monitored for toxigenicity by PCR more
effectively than by other methods such as
microscopy and mouse bioassav.
Baker et al. (2002)
Lake Erie, U.S.
Researchers developed models for predicting
microcystin concentration. Models used 14
physical variables including cyanobacteria DNA,
cyanobacteria RNA, cyanobacterial biovolume
(mass associated with cyanobacteria) and
abundance (number of cyanobacteria per volume of
water). The models were able to accurately predict
microcystin concentration, however, different
factors were important at each site.
Francy and Stelzer
(2014) and Francy et
al. (2015)
Guadarrama
River, Spain
Sequencing of 16S rRNA gene fragments produced
identification of cyanobacteria consistent with
microscopic observations. The 16S rRNA gene is
considered the best target for the phvlogenetic
classification of cyanobacteria and investigating the
di screpancy natural communities of cyanobacteria
(Nubel etal., 1997).
Loza et al. (2013)
Lake Erie, U.S.
Microcystis species could be detected using the
qPCR method.
Wilhelm et al.
(2007)
References are listed at the end of the report.
PCR, polymerase chain reaction.
34

-------
3.3 Design and Siting of Monitoring Networks for Source Water EWSs
Monitoring networks are key elements of a source water EWS. Design of such a network
includes the selection of technology and the siting of monitors. Grayman (2008) lists the
following criteria for the selection of monitoring technology:
•	Cost (capital and operational)
•	Spectrum (broad spectrum or specific constituent)
•	Sensitivity
•	Operational and maintenance requirements
•	Environmental requirements (power, shelter)
•	Sampling frequency
•	Communications requirements
Many studies have addressed the design and siting of monitors within distribution systems.
Murray et al. (2010) provided a summary of the literature in that field with emphasis on U.S.
EPA's Threat Ensemble Vulnerability Assessment - Sensor Placement Optimization Tool
(TEVA-SPOT) software. Hart and Murray (2010) discuss sensor placement strategies for the
design of a distribution system contamination warning system. Though some of the procedures
and characteristics related to monitoring technology developed for distribution system
contamination warning systems are relevant to source water EWSs, the methodologies and
algorithms related to siting monitors in distribution systems are not transferrable to source waters
because of the vast differences between distribution system and surface water configurations.
Strobl and Robillard (2008) reviewed available methods for designing water quality monitoring
networks for surface freshwaters including both monitor locations and monitoring frequency.
However, the methods described in this paper emphasize monitoring networks that are used for
assessing long-term water quality rather than for detecting infrequent contamination and spill
incidents. Only a limited number of studies have been related to siting monitors in natural source
water supplies for detection of sporadic contamination incidents. Grayman and Males (2002)
described a risk-based methodology using a Monte Carlo simulation model for siting monitors
that accounts for the probability of spills, behavior of monitoring equipment, variable hydrology
and the probability of obtaining information about spills independent of a monitoring system.
This model was applied to a 200-mile industrialized stretch of the Ohio River to simulate the
effectiveness of alternative monitoring locations as part of an EWS. Several researchers have
applied optimization algorithms for addressing siting of monitors on rivers to detect spills. Park
et al. (2006), Ouyang et al. (2008) and Telci et al. (2009) used genetic algorithms, and Park et al.
(2010) used an optimization via simulation algorithm with a penalized objective function to
address the water quality monitoring siting problem. In actual practice, monitors are generally
placed in a more ad hoc manner based on "covering" stretches of rivers that are most susceptible
to spill incidents, in the vicinity of important water intakes and locations where the monitors can
be easily serviced and managed.
35

-------
4.0	Statistical Event Detection Methodologies
Online sensors can provide large amounts of water quality data in real-time. In order to be
effective as part of an EWS, statistical mechanisms for quickly evaluating the data and
identifying measurements that might indicate a contamination incident are required. Such
mechanisms are referred to as event detection algorithms and can take many different
mathematical forms. Available event detection methods are reviewed in this chapter.
4.1	Background
Earlier studies of event detection envisioned generation of alarms based on specific or general
measures of water quality crossing thresholds selected based on human health risks or other
measures (Gullick et al., 2003; U.S. EPA, 2005). For example, Gullick et al. (2003) envisioned
an EWS with predetermined response thresholds. Response thresholds were envisioned as set
water quality parameter levels, either absolute or with respect to a baseline, at which a response
is initiated. Responses included confirmation procedures for verifying that excursions beyond
thresholds were real, additional characterization of the incident, characterization of the incident
and assorted response actions. Suggested factors to consider when establishing thresholds
included:
•	Historical patterns of water quality
•	Actual or perceived threat associated with levels of a contaminant or an incident
•	Toxicity of the contaminant being measured
•	Nature and size of the exposed population
•	Ability of treatment processes to remove the contaminant
•	Sensitivity and specificity of the monitoring method
•	Type and severity of action that might be taken when a trigger level is exceeded
Reviews published in the 2000s also noted that thresholds and other parameters should be
adjusted and optimized to minimize false alarms, but still detect credible contamination incidents
(U.S. EPA, 2005). Maintaining a low incidence of false alarms is critical to protecting both
public health and keeping the confidence of the public.
The U.S. EPA (2005) noted the potential for detection of specific contaminants using multiple
water quality parameters. The authors envisioned a contaminant would be detected via a
"signature" discerned from a characteristic pattern of changes in multiple physicochemical
parameters. At the time of the U.S. EPA review, at least one monitoring technology company,
Hach (Loveland, Colorado), had developed a methodology for detecting contaminants in treated
water by integration of data from multiple water quality monitors. As noted below, both
researchers and commercial concerns have continued development of multi-parameter sensors
and data analysis for contaminant detection.
4.2	Integrating Data from Multiple Sensors
Che and Liu (2014) developed and evaluated a relatively straightforward technique for using
multiple water quality measurements (pH, turbidity, conductivity, temperature, oxidation-
reduction potential, UV-254, nitrate-nitrogen and phosphate) to detect specific contaminants
(glyphosate, atrazine, lead nitrate and cadmium nitrate) in treated drinking water. Results of their
36

-------
study should be considered preliminary, since sample water was drawn from a tank with
homogeneous water quality and presence of contaminants did not have to be distinguished from
variations in background water quality. Detection of contaminants was based on correlation
coefficients between pairs of water quality data within a number of time steps (window size).
Under these idealized circumstances, Che and Liu (2014) detected the contaminants used in their
study with a relatively short response time (i.e., time between introduction of the contaminant
and detection of the contaminant) and observed a low rate of false positive observations (i.e.,
determining a contaminant was present when it was not). The choice of window size and
thresholds influenced their proportion of true positive findings. A subsequent study by the same
authors (Liu et al., 2015) extended the approach and based alarms on Euclidean distance of
correlation indicators. The modified method outperformed two other techniques in correctly
detecting actual changes in water quality, though like the former study, experimental work was
not conducted in field conditions and with variable background water quality.
Han et al. (2014) were less successful in detecting specific contaminants using a multi-parameter
sensor approach. The authors conducted their studies in a single-pass pipe loop and injected
organic contaminants (ethylene glycol, 2-methyl-4-chlorophenoxyacetic acid, acetonitrile and
dichloromethane), inorganic contaminants (aqua ammonia and copper sulfate) and simulated
municipal sewage. The system tested was better able to detect nonvolatile organic contaminants
than volatile contaminants and inorganic contaminants, but generally unable to positively
identify any of the detected contaminants. Some of the poor performance noted in the study
related to operational and design problems with the equipment. Those problems included
clogging of instruments by particles and precipitates in the sample water and loss of volatile
constituents in an instruments measuring TOC.
4.3 Event Detection Algorithm Studies
In the past decade, many studies reporting mathematical approaches for detecting incidents from
water quality time series data have been published. Studies span the range of potential
applications (rivers, lake and reservoirs, marine, stormwater, sewage collection, water and
wastewater treatment, drinking water distribution) with the exception of building plumbing
systems. This section overviews the state of the science and focuses on the two most reported
approaches: artificial neural networks (ANNs) and analysis of receiver operating curves.
Two studies (Dawsey et al., 2006; Murray et al., 2011) report the use of Bayesian belief
networks (BBNs) for distribution system contamination event detection. The BBN approach
demonstrated in the studies entails use of a pipe network model (EPANET) to simulate
contaminant injections at various points in a distribution system and characterization of the
responses of sensors installed at various locations in the simulated distribution system.
Potentially, a similar approach using stream network models could be used for event detection in
source waters. The BBNs developed by Murray et al. (2011) successfully identified contaminant
responses in observed experimental data, although the authors also found that the inclusion of
data from unresponsive sensors in their analysis impaired their ability to identify contaminant
responses.
Several studies have reported event detection algorithms based on ANNs. For example, Perelman
et al. (2012) used ANNs to estimate the relationships between water quality parameters in a
treated water distribution system and a Bayesian sequential analysis for estimating the
37

-------
probability of an incident. The authors demonstrated proof-of-concept using a simulated data set
and suggested follow-on research to test, generalize and improve the method's performance.
Wu et al. (2014) conducted a review of the application of ANNs in the field of environmental
and water resources modeling. The review was conducted to characterize ANN modeling
practices and establish best practices (consistency in model development and assurance that
models are sufficiently detailed). ANN model development and application was parsed into six
steps:
•	Input selection (with explicit treatment of significant and independence)
•	Data splitting
•	Model architecture selection
•	Model structure selection
•	Model calibration
•	Model validation
A flowchart showing these steps as a part of a protocol for model development is presented in
Figure 3. The authors conducted critical reviews of 81 published studies on the application of
ANNs for analysis of water quality. Areas of application among the reviewed studies were lakes
and reservoirs (19 studies), rivers (35), groundwater (9), stormwater (4), treatment (8) and
distribution (7). In general, published studies reported similar model development and
application processes. Model architecture selection was judged to be the strongest element of
model development among the studies and input selections was the element requiring the most
improvement. The authors made numerous recommendations for the ongoing development of
ANN models for water quality data analysis, including greater focus on data independence in the
input selection step, better documentation of the model calibration step, and improved
quantification and reporting of uncertainty.
38

-------
Steps in model development process
Method and level of detail
Justification
i j Data collection and pre-processing
f Significance
I 1
| I
_ Independence H~r~j
Input selection
Data splitting <—ฆ
\?
Model architecture selection <—
Method(s) used
and details of
implementation
Reason(s) why
specific method(s)
selected
Model structure selection <-{-ฆ
Model calibration <•—-f-
Predictive <—
Model validation
Structural
Model application I
Figure 3. Proposed protocol for development and application of artificial neural network
(ANN) models (from Wu et al., 2014).
Recently, Oliker and Ostfeld (2015) described inclusion of a pipe network hydraulic model as a
component of an event detection technique. The first step in their event detection technique was
the analysis of each data stream (time series of individual parameter and single locations) via a
"minimum volume ellipsoid (MVE) classifier trained on identifying suspiciously exceptional
measurements" as described in Oliker and Ostfeld (2014). Including network hydraulics and
spatially dispersed sensors resulted in reduced false positive detection rate over event detection
based on single sensors in analyses of a simulated data set. Results of this study hold promise for
event detection including stream network hydraulics for source water event detection.
Several water quality event detection techniques have been included in the CANARY online
water quality data management and event detection tool (McKenna et al., 2008; U.S. EPA,
2012). The algorithms within the CANARY event detection system are based on analysis of
receiver operating characteristic (ROC) curves and include the following options:
39

-------
•	Linear prediction coefficient filter (LPCF)
•	Multivariate nearest neighbor
•	Set-point proximity algorithms
The inclusion of several options for event detection allows CANARY users the opportunity to
base incidents on multiple algorithms (reducing the false positives and negatives) or to determine
the algorithm that provides the most reliable event detection for a particular system or location.
Subsequent to developing CANARY, the U.S. EPA and Sandia National Laboratories developed
and demonstrated techniques for incorporating operational data within distribution system event
detection (Hart et al., 2011). Composite signals (water quality data and other relevant data such
as operational status) can be used to set dynamic event detection set points or for false positive
reduction (i.e., to avoid flagging data as incidents when they can be explained by known
operations actions). Incorporation of operations data is important, since common operations such
as tank emptying and filling can produce water quality signals that appear as incidents. An
alternative approach to ensuring operations and routine water quality changes are not interpreted
as real incidents was demonstrated by Zhao et al. (2015). In that study, patterns in historic data
were recognized and used in interpretation of water quality data. The authors found that for the
tests they performed, interpreting data in the light of routine patterns allowed more sensitive
event detection as well as a reduction in the number of false positive incidents. An analogy can
be drawn between water quality changes driven by operation of distribution systems and water
quality driven by precipitation in source water. In both cases, an external stimulus known to the
system operator causes a rapid, but expected water quality change.
40

-------
5.0	Modeling as an Element of Early
Warning Systems
Mathematical modeling of water quality in surface water dates back to pioneering work on the
oxygen balance in the Ohio River (Streeter and Phelps, 1925). With the advent of digital
computers, computer models were developed to simulate surface water quality (Thomann, 1963).
Subsequently, extensive development and application of computer models to simulate surface
water quality has occurred. Water quality modeling has been used for a variety of activities such
as: development of discharge permits, waste load allocations, impact assessment of non-point
sources and combined sewer overflows and, most relevant to EWSs, prediction of movement and
impacts of transient spills.
5.1	Background
Grayman et al. (2001) presented an in-depth review of surface water models for use in EWSs.
The following material is drawn directly from that report and serves as a review of the state of
the science of modeling as an element of EWSs circa 2001.
•	Spill models are a class of models that are used to trace the movement and fate of
transient contaminants in receiving water. They are generally used in real-time or near
real-time situations.
•	Streams and rivers, lakes and reservoirs and the non-saline portions of estuaries are
potential sources of drinking water and thus, are the surface water categories that can
benefit most greatly from spill models as they apply to drinking water sources.
•	Three basic components of any spill model include: a flow module, a water quality
transport module and a fate module. The flow module describes the movement of the
water; the water quality transport module describes the processes by which the
contaminant concentration moves and changes due to the hydrodynamic forces; and the
fate module describes the impacts of physical, chemical and biological processes on the
form and concentration of the contaminant. These modules could be represented in
separate models that are interconnected through input-output or might be integrated into a
single model that represents the entire process.
•	Flow models used in EWS modeling can be classified as: non-hydraulic methods,
hydraulic methods for steady flows, unsteady flow models, lake and reservoir models and
estuarine models. Each of these general methods has a potential place in spill models.
•	Water quality transport models represent the movement of contaminants in the aquatic
environment. Transport processes include the advective movement and the diffusive
processes that spread the contaminant. Diffusive processes include both molecular
diffusivity and turbulent diffusion. In many situations, models can be used that are
simplifications to the full three-dimensional equations. The most common simplifications
are averaging over one or two dimensions.
•	Fate models that are part of real-time EWS models generally fall into one of three
categories: representation of substances as conservative, use of a simple decay model,
and development of fate parameters and processes for the most commonly encountered
substances.
41

-------
•	A range of alternatives in terms of incorporating a model into an EWS include:
development of a special purpose model, modification of an existing model, and use of an
existing model without modification.
•	A separate class of models were identified that modeled inland oil spills on rivers.
Physical-chemical processes that affect fate and transport in these types of models
include: advection due to wind and current, spreading of surface oil due to turbulent
diffusion and mechanical spreading, emulsification and spreading of oil over the depth of
the river, changes in mass and physical/chemical properties due to weathering, interaction
of oil with the river shore lines, attachment of oil droplets to suspended particulates,
photo-chemical reactions, and microbial biodegradation.
Grayman et al. (2001) summarized the state of development and applications of models as part of
EWS at that point of time and highlighted the following observations:
•	Most model development by 2001 had targeted planning and impact analysis
applications; relatively little effort had been spent on development of models for early
warning.
•	Most model development had been done for very large rivers in the U.S. and Europe.
Lake and reservoir modeling is more complex than riverine models.
•	A critical component of source water EWS modeling is a mechanism for determining
flow conditions and for predicting contaminant fate and transport. An EWS could
incorporate these elements via internal tools or via link to external tools and data sources.
5.2 Taxonomy of EWS Models
Models used as part of EWSs can be categorized based on complexity, processes represented,
availability of models, and model design. For the purpose of this review, the following four
general categories of models have been selected:
•	Physically based models: models that represent the underlying physical-chemical
processes
•	Geographic information system (GlS)-based models: a variety of process models that
have been integrated with a geographic information system for ease in parameterizing the
model and displaying results
•	Data-driven models: models that rely upon analysis of large databases of observed data
(input-output models) rather than emphasizing the underlying physical processes
•	Simplified modeling techniques: models that have used highly simplified representations
of the underlying physical-chemical processes
It should be noted that these categories are not necessarily unique and a particular model could
fit into multiple categories (i.e., a physically-based model or a simplified model could be
integrated within a GIS). In the following sections, progress in EWS modeling is discussed in
terms of the four categories.
5.2.1 Physically based Models
Physically-based hydraulic and water quality modeling is a well-developed field that dates back
over half a century. Grayman et al. (2001) provided a picture of the state of the science of
physically based modeling in 2001. Wang et al. (2013) provided a review of water quality
42

-------
models for surface water as of 2013 while Bahadur et al. (2013) presented a state of the science
review of water contamination modeling as of 2013. Advances in the field made over the past 15
years are described below.
HEC-RAS is the widely used USACE's Hydrologic Engineering Center model designed to
perform one-dimensional hydraulic calculations for a full network of natural and constructed
channels. It is currently being expanded to include riverine water quality analyses (USACE,
2016). Transport and fate of a limited set of water quality constituents is now available in HEC-
RAS. The currently available water quality constituents are: dissolved nitrogen (NO3-N, NO2-N,
NH4-N and Org-N), dissolved phosphorus (PO4-P and Org-P), algae, DO, carbonaceous
biological oxygen demand (CBOD) and water temperature. Testing of this module and further
expansion of its capabilities is currently underway. This addition to HEC-RAS is significant
because of the wide use of the hydraulic model. Simultaneous hydraulic and water quality
modeling allow the use of HEC-RAS within an EWS for prediction of times of travel for
contaminant plumes and estimation of contaminant attenuation during transport.
The MIKE series of models (MIKE11, MIKE21, MIKE3) are three physically based models
developed by the company DHI (Horsholm, Denmark) for simulating flow and water quality in
one, two and three dimensions, respectively. This series of commercial modeling systems are
widely used worldwide and have been implemented as part of an EWS on the Yellow River in
China (Burchard-Levine et al., 2012).
Wang et al. (2013) deemed the period after 1995 as the "deepening stage" in surface water
quality modeling. Advances that they identified during this period include: improvements in
modeling non-point sources and inclusion of nutrients and toxic chemical materials depositing to
water and land surfaces. The authors identified seven groupings of models with varying
complexity and data requirements. They also noted a trend towards standardization of models
and their application as evidenced by the U.S. EPA's guidance for quality assurance programs
associated with models (U.S. EPA, 2002). Wang et al. (2013) called for more standardization
and greater attention to validation during model development. In addition, the authors noted that
models needed to be developed for the appropriate use.
Bahadur et al. (2013) conducted a state of the science review of water contamination modeling.
In the paper, 65 separate models were identified and categorized in terms of environment (river,
lake, estuary, coastal, watershed), degree of analysis (screening, intermediate, advanced),
availability (public, proprietary, restricted support), temporal variability (steady state, time
variable), spatial resolution (1-D, 2-D, 3-D), processes (flow, transport, both), water quality
(chemical, biological, radionuclide, sediment), and support (user support, use manual). Based on
the dates of the references given in the report, less than 25% of the models were developed or
significantly updated after 2001. An examination of the post 2001 models mentioned in the
review included HEC-RAS, MIKE, RiverSpill and ICWater (Incident Command Tool for
Drinking Water Protection) models as the most relevant in terms of modeling as part of EWSs.
These models are discussed elsewhere in the present report.
5.2.2 GIS-based Models
GIS dates back to the 1960s and 1970s (McHarg, 1969; Tomlinson, 1968). In the 1980s, Horn
and Grayman (1993) introduced the concept of integrating water quality modeling into the
nationwide riverine-based Reach File System for planning studies. In a demonstration project for
the U.S. EPA, Grayman et al. (1994) integrated a simplified steady-state water quality model
43

-------
with the Reach File System using a commercial GIS package (ARC/INFO™ NETWORK [ESRI,
Redlands, CA]). The system was implemented for the Ohio River Basin and could be used to
calculate the path that a spill at any point in the basin would follow, and the resulting
concentrations and travel time throughout the downstream path.
Since 2001, several GIS-based spill models have been proposed and implemented. A consortium
of federal agencies sponsored the development of RiverSpill software, a GIS-based software
package that calculates time-of-travel and concentration of contaminants in streams and rivers
(Samuels et al., 2006). RiverSpill software uses real-time stream flow data, a hydrologically
connected stream network (USGS Enhanced Reach File version 2.0 ERF1-2) and the locations
and populations served by each public, surface drinking water intake. It could be applied to
simulate deliberate contamination incidents or accidental water contamination incidents, such as
spills from transportation accidents on roadways and railroad, pipelines, wastewater treatment
plants and hazardous materials storage sites. ICWater (Incident Command Tool for Drinking
Water Protection) software evolved from RiverSpill software and provides real-time assessments
of the travel and dispersion of contaminants in streams and rivers (Samuels et al., 2015). It uses
the 1:100,000 scale National Hydrography DatasetPlus, Version 1.0 (NHDPlusVl), a
hydrologically connected river network that contains over three million reach segments in the
United States. Mean flow and velocity have been calculated by the USGS and the U.S. EPA for
each reach. These mean values are updated by flow from web accessible real-time gauging
stations. ICWater software was successfully applied in near real-time during the 2014 Elk River
spill incident to predict the movement of the spill in the Ohio River (Bahadur and Samuels,
2015).
Zhang et al. (2011) developed a GIS-based spill model for the Huaihe River Basin that runs
through central and eastern China. The system includes database management, water quality
evaluation, statistical analysis, case management, model simulation, and emergency response as
features of an integrated water-pollution emergency-information-management and decision-
making system for river-basin management.
Rui et al. (2015) proposed an emergency response system based on the integration of GIS
technology and a hydraulic/water-quality model. Using the spatial analysis and three-
dimensional visualization capabilities of GIS technology, they calculated pollutant diffusion
measures, and visualized and analyzed the simulation results. The system has been demonstrated
as part of an early warning and emergency response program for sudden water pollution
accidents in the Xiangjiba Dam area on the Yangtze River in China.
Significant interest exists in applying GIS and modeling for oil spill response and analysis. The
General NOAA Oil Modeling Environment (GNOME) has been combined with remote sensing
by National Aeronautics and Space Administration (NASA) to assess the oil spill modeling
potential (Spruce, 2004). Chen et al. (2010) developed a two-dimensional hydrodynamic/oil spill
model based on GIS to simulate currents and oil transportation in rivers, lakes and reservoirs,
and applied it to the Three Gorges Reservoir in China.
GIS-based spill models used in conjunction with national databases have the potential for
significantly reducing the setup time and the effort required to implement a spill model. Unlike a
conventional stand-alone spill model that requires extensive manual parameterization and
calibration and thus, must be in place prior to an actual spill incident, a GIS spill model (in
conjunction with an available national database) could be quickly implemented. At the extreme,
44

-------
such a system could be implemented after a spill is detected and used as a first-cut predictive tool
to simulate the incident in near real-time. The application of ICWater model to the Elk River
spill on a near real-time basis demonstrated this potential. However, establishment and testing of
any spill modeling system on a river of interest prior to an actual incident will increase the
confidence associated with the resulting simulations.
Though the promise of GIS as a component of EWSs is clear, numerous challenges exist in
regards to its effective use in this context. First and foremost, GIS, as an information system, has
significant data needs ranging from sufficiently detailed representations of stream networks,
lakes and reservoirs to locations of contaminants or spill sites. The dynamics of land use,
hydrology, climate and other temporally varying factors require update and maintenance of GIS
data. Second, for its most effective use in an EWS, GIS should be integrated with tools such as
event detection systems (inclusive of means for updating and maintaining water quality time
series data) and hydraulic models. Although commercial and open source GIS packages have
facilities for integrating EWS tools with GIS, integrated tools are not currently well-tested or
readily available. Finally, data requirements as well as upkeep of GIS tools themselves can
require dedicated staff and significant expenses.
5.2.3 Data-driven Models
Whereas physically based models are based on mathematical descriptions (i.e., the physics) of
river flow and water quality transport/fate, data-driven models (DDM) rely upon various
methods that analyze data sets to draw conclusions related to the nature of the problem (flow,
transport and fate). Solomatine and Ostfeld (2008) described the DDM process as: "a
dependence ('model') is discovered (induced), which can be used to predict the future system's
outputs from the known input values" - i.e., an input-output model.
Burchard-Levine et al. (2012) identified the three most frequently used data-driven models as
statistical, fuzzy logic and ANN, with ANN currently being the most widely used technique in
the area of water modeling. Maier et al. (2010) reviewed 210 journal papers that were published
from 1999 to 2007 and focused on the use of ANN in the prediction of water variables in river
systems. Approximately 90% of the applications focused on water quantity (flow and level) with
only 10% applying the method to predicting a wide range of water quality variables. The review
did not explicitly indicate those studies that pertained to EWS modeling, but an examination of
the reference list suggested very little research work that has been directly applied to early
warning modeling.
Piotrowski et al. (2007) described a new ANN-based approach that relies heavily on
measurements of concentration collected during tracer tests over a range of flow conditions to
develop a predictive capability. Four separately designed neural networks were used to predict
concentration versus time measurements at a particular cross-section as characterized by the peak
concentration, the arrival time of the peak at the cross-section and the shapes of the rising and
falling limbs.
In probably the most applicable study in the use of DDM to EWSs, Burchard-Levine et al.
(2014) performed a case study in a southern industrial city in China in which a DDM based on
genetic algorithms (GAs) and ANN was tested. The GA-ANN model was used to predict NH3-N,
chemical oxygen demand (COD) and TOC variables at a downstream station two hours ahead of
time resulting in an increase in the response time of the city's EWS.
45

-------
Deng (2014) developed a modeling capability as part of a pollution EWS for the management of
water quality in oyster harvesting areas along the Louisiana Gulf Coast. The system consists of
(1) an Integrated Space-Ground Sensing System gathering data for environmental factors
influencing water quality, (2) an ANN model for predicting the level of fecal coliform bacteria
and (3) a web-enabled, user-friendly GIS platform for issuing water pollution advisories and
managing oyster harvesting water.
5.2.4 Simplified Modeling Techniques
The fourth category of models utilizes simplified relationships and methods to predict the
movement of contaminants in rivers. Many of the first riverine EWSs employed such methods.
Fennell (1988) described a simplified time-of-travel and peak concentration model developed by
ORSANCO in response to a major oil spill in 1988. Other simplified models were applied to the
Rhine River (Spreafico and van Mazijk, 1993) and the Lower Mississippi River (Waldon, 1998).
Simplified models typically assume one-dimensional steady flow, calculate the relationship
between velocity and flow by a simple power function equation and either ignore or use
simplified methods for representing dispersion and decay. In many cases, the simplified methods
utilize spreadsheets as the calculation mechanism.
Since 2001, minimal use and development has occurred in the area of simplified modeling for
EWSs. The GIS-based models described earlier (RiverSpill and ICWater software) utilize
modeling methods that could be best described as enhanced simplified modeling that fall
between simplified models and physically based models.
46

-------
6.0	EWS Data Integration and Communication
Previous chapters in this report have presented information on specific components of EWSs
including monitors, modeling and event detection. However, in order for there to be a 'system' in
EWSs, linkages between the individual components are needed. U.S. EPA developed guidance
on designing a real-time source water quality monitoring system to detect source water
contamination incidents that provides information on data integration and communication (U.S.
EPA, 2016c). In this chapter, three forms of linkages are discussed and past work in these areas
are reviewed: data integration, communications and institutions.
6.1	Data Integration
EWS data can include:
•	Water quality data (data input to and managed by an EW S)
•	Other data that could indicate a water quality incident of significance to drinking water
operations and their customers (e.g., spill reports, CSOs, incident notifications)
•	Contextual data such as spill locations, drinking water treatment operation intake
locations, zones of critical concern, watershed boundaries, locations of potential
contaminants of concern (e.g., storage tank locations)
•	Interpretive data such as alarms and event detection application outputs and inputs
•	Incident response data such as actions taken in response to an incident (including
classification of an anomalous water quality change as a non-incident), results from
sampling in response to an incident
Although an EWS does not necessarily house and use all of those data, access to all of them
could facilitate improved incident detection and response.
The data described above are likely to come from a variety of agencies (e.g., U.S. EPA, state
regulatory agencies, public water system monitoring data, USGS) and differ in their structure,
completeness and use. Ideally, an EWS would have the ability to integrate those data in a tool
that enables their effective use and maintenance. Such a tool would give users a unified view of
the data and means for easily querying other pertinent data.
Three software tools have been developed for data integration as related to source water
protection, spill response and incident detection. Each of these systems utilizes a GIS platform to
display spatial data and integrates information derived from public databases and other data
sources. The three tools are described below.
•	The Drinking Water Mapping Application for Protecting Source Waters (DWMAPS)
application is an internet-based GIS tool for drinking water source water protection and
assessment (U.S. EPA, 2016a). The DWMAPS tool is currently under active
development for the U.S. EPA's Office of Ground Water and Drinking Water. It includes
a mapping tool, a linked source water protection planning tool and a suite of data
exchange services that could be used to display and assess contextual data and identify
potential contamination sources.
•	The ICWater tool is a GIS-based tool developed for a consortium of federal agencies (see
the section on GIS-based models in chapter 5 for additional information on this system).
47

-------
The ICWater tool is linked to national databases containing information on dams,
reservoirs, water supplies, gauges, municipal and industrial dischargers, and
transportation networks. It integrates a riverine contamination model and the
aforementioned national databases, within a near real-time response framework.
• WaterSuiteฎ software (Rockland, MA) is a suite of tools integrated into a GIS framework
(Rosen and Kearns, 2015). Tools currently in place within WaterSuite software are a
source water assessment and protection tool, a water quality data management and event
detection tool, a distribution system data management and analysis tool and a drinking
water treatment database. Source water protection data housed in WaterSuite software
include federal, state and local data sets with features related to potential sources of
contamination, a contaminants database and data created and managed by utilities using
the tool. At present, the water quality data management and event detection module pulls
georeferenced, time-stamped online water quality data from sensors, houses those data in
an efficient data structure and displays those data in a user interface. Near-future
developments of the water quality module include integration of event detection software.
In a related area, the (U.S.) Federal Geographic Data Committee and the Advisory Committee on
Water Information launched the Open Water Data Initiative in the summer of 2014 (Rea et al.,
2015). The goal of the initiative is to bring currently fragmented water information into a
connected, national water data framework by leveraging existing systems, infrastructure and
tools to underpin innovation, modeling, data sharing and solution development. As part of this
effort, a workgroup consisting of representatives from the U.S. EPA, U.S. Department of
Homeland Security (DHS), USGS, NOAA, private industry and academia has been formed to
investigate existing applications that address modeling and simulation, web services, GIS
mapping, hydrology, emergency response, exercises and contingency planning.
6.2 Communications
Communications is a significant aspect of EWSs that includes (1) communication between
sensors and a central control point, (2) communication between the central control point and a
user interface or other data visualization and analysis tool, (3) communication between agencies
and (4) communication between agencies and the public. Ideally, communication between
agencies and agencies and the public is two-way and is conducted through the most appropriate
channels for each stakeholder group. Communication of EWS data is done in the context of
overall incident response communication and coordination. The communication could include
the involvement of multiple cities, states, or other structures like incident command operations
external to the EWS. Effective communication of EWS technical information could pose greater
challenges than collection and analysis of the technical information and can be a focus for future
EWS research and development. U.S. EPA provides guidance and tools on the topics of
communications technologies, information management, consequence management, and risk
communication for water quality surveillance and response systems that could be useful for
source water EWSs at https://www.epa.gov/waterqualitvsurveillance/water-qualitv-surveillance-
and-response-svstems-guidance-and-tools.
Panguluri et al. (2005) described the options for communication between field-based sensors and
a central control point. The authors listed the factors influencing the selection of communication
technologies as availability, cost, user preference, and the relative location of the sensors to the
data acquisition system. Additional factors the authors have encountered when establishing
48

-------
communication between field-based sensors and a central control point include accessibility
(including strength of cellular signal, if that mode is selected, and access to a power supply) and
security concerns (e.g., isolating communications from critical network infrastructure).
Communication can be conducted either by wired (e.g., direct, phone line) or wireless (e.g.,
radio, cellular, satellite, WiFiฎ, ZigBeeฎ, Bluetoothฎ). Campisano et al. (2013) described real-
time control of urban wastewater systems and details on the communication options that are
available for collecting water quality data and controlling the wastewater collection system.
Communication options include phone lines (leased or dial-up), ethernet networks over fiber
optic or copper cables (private or leased networks), cellular data communication services and
radio communication networks (licensed or unlicensed). Radio communication can be effective
for communication over relatively short distances and has been demonstrated during prior water
quality monitoring studies (e.g., Anvari et al., 2009; Dehua et al., 2012; Glasgow et al., 2004;
Jiang et al., 2009). U.S. EPA provided guidance on designing communication networks for water
quality monitoring systems (U.S. EPA, 2016b).
The National Response Center (NRC) is the federal government's national communications
center, which is staffed 24 hours a day by USCG officers and marine science technicians. The
NRC is the sole federal point of contact for reporting all hazardous substances releases and oil
spills. The NRC receives all reports of releases involving hazardous substances and oil that
trigger federal notification requirements under several laws. Reports to the NRC activate the
National Contingency Plan and the federal government's response capabilities. It is the
responsibility of the NRC staff to notify the pre-designated on-scene coordinator assigned to the
area of the incident and to collect available information on the size and nature of the release, the
facility or vessel involved and the party(ies) responsible for the release. The NRC maintains
reports of all releases and spills in a national database.
For specific localities and types of spills, a variety of mechanisms have been developed to help
communicate and manage spills. Two examples include:
•	The Metropolitan Water Reclamation District of Greater Chicago (MWRDGC)
coordinates with the City of Chicago and all suburban municipalities to notify suppliers
of potable water of contamination from CSOs. MWRDGC has created a web page
(http://www.mwrd.org/irj/portal/anonymous/overview) on their website to inform the
general public of the occurrences of CSOs. A color-coded graphic representation of the
waterways appears on the web page depicting the occurrence of CSOs and waterway
diversions to Lake Michigan.
•	The Metropolitan Sewer District of Greater Cincinnati and its partners, Sanitation District
No. 1 of Northern Kentucky (SD1) and ORSANCO have developed the Recr80hioRiver
website and an associated software application. It provides Ohio River water quality
information and river conditions in the Greater Cincinnati area to recreational users and
others as they are planning to boat, fish, swim, or engage in other water sports
(http://www.recr8ohioriver.org/default.aspx) (accessed January 19, 2016).
6.3 Institutions
49

-------
Based on the case studies documented in this report, a variety of institutional structures are
available for managing spills. These include public regional agencies, industry groups and utility
managed systems (individual utilities or consortiums).
Grayman et al. (2014) discussed institutional issues as follows:
•	Multiple agencies and institutions might be involved in the detection, co-
ordination and mitigation of a spill incident. Since rapid response is generally
important in dealing with a spill, pre-planned institutional responsibilities,
protocols and arrangements are needed.
•	Effective spill response requires a lead organization to serve as the overall
coordinator during emergency incidents. The lead agency can be a regional
agency, a governmental unit or a water utility. Examples of such organizations
include international agencies (Rhine River), state agencies (Louisiana), federal-
state commissions (ORSANCO, Ohio River), water utilities (UK), a group of
water agencies (Japan) and private organizations (St. Clair River).
•	Other agencies that could be involved in spill situations in the U.S. include the
NRC, the USCG, state and federal environmental agencies, US ACE, local health
and environmental agencies, emergency responders, law enforcement
organizations and water utilities. Effective interaction among the agencies is a key
to successful operation during a spill situation.
50

-------
7.0 Conclusions and Research Needs
Based upon the state-of-the-science review of research on EWSs, several conclusions and needs
have been identified relative to the research, development and establishment of source water
EWSs.
7.1 Conclusions
Interest in EWSs for addressing and responding to contamination incidents in surface water
systems has recently increased as a result of several high profile incidents. Contamination
incidents might be caused by a wide range of sources including industrial and transportation
related spills, non-point sources and urban runoff, intentional contamination, and natural
processes. EWSs encompass much more than just monitors; rather they include mechanisms for
detecting, characterizing, communicating and responding to contamination incidents in order to
reduce and mitigate the impacts. Lak Though commonalities exist between source water EWS
and contamination warning systems in drinking water distribution systems, there are significant
differences in terms of the sources of contamination, the configurations of the receiving water
infrastructure (pipes versus rivers/lakes), the characteristics of the water medium (raw versus
finished water), the response protocols, and the methods of selecting and siting monitors to
detect the presence of contaminants.
Following the terrorist incidents on September 11, 2001, emphasis in the area of EWSs shifted to
potential intentional contamination of water distribution systems resulting in a robust research,
development and implementation program for contaminant warning systems in drinking water
distribution systems. In the last quarter of the twentieth century, a significant number of EWSs
for source waters were established around the world, in most cases, in direct response to
contamination incidents. A chemical spill into the Elk River in West Virginia in 2014 followed
by several other much publicized spills has reinvigorated the interest in source water
contamination EWSs. In addition, the threat of HABs in drinking water source water has
emerged in recent years as a significant challenge to the drinking water community. Early
warning activities focused on detecting HABs differ significantly from most water quality
contaminants and additional research is needed in this area.
New online water quality monitoring technologies are emerging for use in EWSs. Online
technologies automatically collect and communicate data and monitor a flowing sample or
collect and analyze discrete samples. Emerging technologies vary widely in their state of
development (from conceptual to commercialized), the focus on contaminants of relevance to
drinking source water, and the potential for field deployment. The most relevant emerging
technologies to drinking source water EWSs are biomonitors (monitors using the response of
biological organisms to water constituents) and spectroscopic instruments (instruments sensing
absorbance, transmittance and scattering/reflectance of electromagnetic radiation).
Several criteria should be used in selecting monitoring technology for use in an EWS: cost,
spectrum (broad spectrum or specific constituent), sensitivity, operational and maintenance
requirements, environmental requirements (power, shelter), sampling frequency, and
communications requirements. Though many studies have addressed the design and siting of
monitors within distribution systems, only limited research has been done into siting monitors in
source waters. In actual practice, monitors are generally placed in a more ad hoc manner based
51

-------
on "covering" stretches of rivers that are most susceptible to spill incidents, in the vicinity of
important water intakes and locations where the monitors can be easily serviced and managed.
Similarly, many event detection methodologies and techniques have been applied successfully to
drinking water distribution systems, but their application to source water event detection has
been limited. Source water quality is much more variable than water quality in distribution
systems, making anomalous water quality incidents more difficult to discern from background
variations. Additionally, rainfall and runoff are regular occurrences for source water and it is
unclear how event detection methodologies will account for water quality changes associated
with these common occurrences that sometimes would not be considered contamination
incidents. Some initial application studies are promising, but additional work is needed to
determine the applicability of existing technologies to surface water EWS. If techniques
developed for drinking water distribution system event detection do not perform as well for
source water incident detection, it is possible that techniques such as artificial neural networks,
correlations between multiple water quality parameters, or development libraries of user-defined
signatures of non-incidents could be used to modify existing algorithms for better application to
source water incident detection.
Spill models are a class of models that are used to trace the movement and fate of transient
contaminants in receiving waters. They can be used in an EWS in real-time or near real-time
situations to assist in the response to a spill or other transient contaminant incidents or could be
used in historical reconstruction of past incidents. Models have been categorized as physically
based models, GIS-based models, data-driven models and simplified modeling techniques. Prior
to 2000, most models were physically based models, simplified models or rudimentary GIS-
based models. Since 2000, development and research in spill modeling has emphasized data-
driven models and GIS-based models. Data-driven models utilize artificial neural networks,
fuzzy logic, statistical methods and other methods to model flow and water quality in rivers.
GIS-based models integrate models with GIS and national databases, such as the USGS National
Hydrography Dataset, and facilitate EWS modeling with minimal setup and parameterization.
EWSs are composed of the individual components of monitoring, modeling and incident
detection. However, in order for an EWS to function effectively, linkages should exist between
the individual components. These linkages include data integration, communications and
institutions. A single platform to integrate data from the individual components can make an
EWS more efficient.
7.2 Research Needs
To better understand the performance, detection capabilities and limitations of source water
EWSs, more research needs to be conducted. One important aspect of identifying research needs
is to conduct more assessments on existing EWSs. The assessments could provide data on the
type of contamination incidents that are routinely detected and the types of contamination
incidents that are not detected by EWS. More information on the effectiveness of an EWS could
be completed by conducting a simulation study to assess the likelihood of detecting a broad set
of potential contamination incidents given different types of EWS. Additional future research
needs identified as part of this study are in the areas of contaminant information, monitoring
technologies, placement of monitors, event detection methodologies, fate and transport models
and data management and visualization tools.
7.2.1 Contaminant Information
52

-------
To improve the effectiveness of a source water EWS, more information is needed on the
contaminants that might be a possible threat. The research needs associated with this involved:
•	Developing tools that enable better access to contaminant information in the watershed,
such as inventories of shipping cargo for rail cars, barges and trucks, storage tanks near
surface water, and pipelines crossing surface water bodes
•	Evaluating the need for this information to be available in real-time, spatially and
temporally, in order to be linked to surface water modeling and simulation tools
Determining kinetic, partitioning information, and reaction dynamics for the contaminants that
are of greatest concern
7.2.2	Monitoring Technologies
In terms of the monitoring technologies used for source water early warning systems, research
needs identified in the development of this report include:
•	Determining which parameters to monitor and how they relate to the contaminants of
greatest concern to a drinking water utility or downstream use
•	Assessing the data available from existing EWS to better understand the field
performance of various monitoring technologies
•	Evaluating monitoring technologies through bench, pilot and field scale testing for
emerging contaminants and harmful algal blooms, such as:
o Biomonitors
o Spectroscopic instruments
o Fluorescence/spectral instruments
o Hydrocarbon detection instruments
o Multi-parameter probes
o Optical monitors (e.g., DO)
o Detailed organic matter monitors (related to DBP formations)
o Solid state and lab-on-a-chip technologies
•	Developing monitoring technologies that are:
o More reliable real-time instruments to operate in source water with varying water
quality
o More practical and accurate for inorganic contaminants
7.2.3	Placement of Monitors
More research needs to be completed to help identify where monitors should be located with the
source water to be the most effective for the purposes of the early warning system. This could
involve the following activities:
•	Developing approaches and tools that incorporate upstream threat analysis to go beyond
the traditional approach of siting monitors near an intake or just downstream of a known
threat. Upstream threat analysis and identification of monitoring locations should
consider the area tens to hundreds of miles upstream of a drinking water intake to account
for variations in flow rates, volumes, speeds and contaminant characteristics (e.g.,
conservative, volatility, density), since time-of-travel can vary from days to weeks with
different downstream peak concentrations.
53

-------
•	Evaluating the siting of monitoring locations as part of resilience planning. Monitoring
locations could be selected in such a way as to improve the utility's ability to react and
respond to spills and to plan for long-term capital improvements (asset management).
Resilience planning can include both long term chronic risks (lower concentration) and
acute spills (higher concentrations) to evaluate the utility's ability to counter different
threats by:
o Intake closure/finished water storage
o Treatment (emergency or permanent change to the treatment train)
o Alternative sources of raw/finished water
•	Evaluating the placement of monitoring locations in regards to how they might be used to
enhance daily operational controls for large reservoirs on tributaries and on navigational
locks and dams. Water quality/volume can be accounted for in multiple-use reservoir
flow releases for flood and water quality management, and how they impact downstream
water quality and navigational needs.
•	Evaluating the siting of monitoring locations based upon a contamination scenario, in
which multiple individual worst case scenario emergency response plans are combined.
An example scenario could be a chemical tank that ruptures and spills into the river, and
then, during the response, a barge carrying a different chemical compound wrecks and
spills its content into the river as well.
7.2.4	Event Detection Methodologies
Additional research needs are associated with the application of EDS to source water
applications. These needs could include:
•	Assessing the data available from existing EWS to better understand current false
positive detection rates and what causes them
•	Evaluating the existing EDS tools and approaches on water from a variety of surface
water bodies to develop correlations between multiple water quality parameters and
contaminants of concern
•	Developing libraries of events/alarms associated with common contaminants and user-
defined signatures of non-incidents to help reduce the high frequency of false positive
assessments
•	Developing additional EDS techniques that could explore the use of artificial neural
networks
7.2.5	Fate and Transport Models
Another research area identified in this study is related to fate and transport models that could be
used to support source water early warning systems, including:
•	Developing approaches to link real-time data (e.g., flows, levels, water quality
parameters) with surface water modeling and simulation tools to obtain a more accurate
representation of the water system
•	Developing modeling and simulation tools that incorporate the whole watershed, such as
overland flows, surface water flows, wastewater discharges, and drinking water
distribution systems
•	Developing linkages between the fate and transport models and GIS databases which
contain information about the type of contaminants that might be in the watershed
54

-------
•	Incorporating cross-sectional stream geometries, if available, into three-dimensional
surface water models to support more accurate fate and transport of spills
•	Evaluating the need to verify and calibrate three-dimensional models to determine if the
accuracy is necessary to support effective source water early warning systems
7.2.6 Data Management and Visualization Tools
In addition, a source water EWS requires data management and visualization tools to support
analytics and communication. Some research needs identified in the study include:
•	Developing better data transmission tools to support monitor at remote sensing locations
•	Evaluating alternative power supply options to remote sensing locations
•	Developing enhanced data analysis and visualization tools to support real-time response
actions
•	Developing a reliable method for validating data from online instruments in real-time
•	Evaluating vulnerabilities in the data communication pathway, including cybersecurity
concerns
•	Evaluating the benefits of sharing data with a wider set of stakeholders
•	Evaluating techniques to incorporate reporting of contamination incidents from social
media, citizen science, or other public sources
55

-------
8.0 References
Allen, J.B., Franz, B., Hokanson, D., Panguluri, S., Carson, J., 2014. Source water monitoring
and biomonitoring systems. Presented at Source Water Contaminant Detection Training:
Early Warning and Response, Morgantown, WV.
Anderson, K., 2015. The Delaware Valley early warning system. 2015 Exchange Network
National Meeting, Philadelphia PA.
http://www.exchangenetwork.net/EN2015_files/9.3.pdf
Anvari, A., Delos Reyes, J., Esmaeilzadeh, E., Jarvandi, A., Langley, N., Navia, K.R., 2009.
Designing an automated water quality monitoring system for West and Rhode Rivers, in:
Proceedings of the 2009 IEEE Systems and Information Engineering Design Symposium.
Presented at the IEEE Systems and Information Engineering Design Symposium,
University of Virginia, Charlottsville, VA.
Bahadur, R., Amstutz, D.E., Samuels, W.B., 2013. Water contamination modeling—A review of
the state of the science. J. Water Resour. Prot. 5(2): 142-155.
Bahadur, R., Samuels, W., 2015. Modeling the fate and transport of a chemical spill in the Elk
River, West Virginia. J. Environ. Eng. 141(7):5014007.
Baker, J. A., Entsch, B., Neilan, B.A., McKay, D.B., 2002. Monitoring changing toxigenicity of a
cyanobacterial bloom by molecular methods. Appl. Environ. Microbiol. 68(12):6070-
6076.
Banna, M.H., Imran, S., Francisque, A., Najjaran, H., Sadiq, R., Rodriguez, M., Hoorfar, M.,
2014. Online drinking water quality monitoring: review on available and emerging
technologies. Crit. Rev. Environ. Sci. Technol. 44(12): 1370-1421.
Bode, H., Nusch, E.A., 1999. Advanced river quality monitoring in the Ruhr basin. Water Sci.
Technol., Management of Large River Basins Selected Proceedings of the 8th IAWQ
River Basin Conference 40(10): 145-152.
Bowling, L., Ryan, D., Holliday, J., Honeyman, G., 2013. Evaluation of in situ fluorometry to
determine cyanobacterial abundance in the Murray and Lower Darling Rivers, Australia.
River Res. Appl. 29(8): 1059-1071.
Bridgeman, J., Baker, A., Brown, D., Boxall, J.B., 2015. Portable LED fluorescence
instrumentation for the rapid assessment of potable water quality. Sci. Total Environ.
524-525:338-346.
Brosnan, T.H. (ed.), 1999. Early warning monitoring to detect hazardous events in water
supplies, ILSI Risk Science Institute Workshop Report, Washington DC.
Burchard-Levine, A., Liu, S., Vince, F., 2012. Drinking water source contamination early
warning system and modelling in China: A review. Int. J. Environ. Pollut. Remediat.
1:13-19.
Burchard-Levine, A., Liu, S., Vince, F., Li, M., Ostfeld, A., 2014. A hybrid evolutionary data
driven model for river water quality early warning. J. Environ. Manage. 143:8-16.
Cagnard, O., Baudin, I., Lemoigne, I., Cartnick, K., 2006. Assessment of emerging optic sensors
(fluoroprobes) for algae on-line monitoring. Water Quality Technology Conference and
Exposition, Denver, CO.
Campisano, A., Pie, J.C., Muschalla, D., Pleau, M., Vanrolleghem, P.A., 2013. Potential and
limitations of modern equipment for real time control of urban wastewater systems.
Urban Water J. 10(5):300-311.
56

-------
Carmichael, W., 2000. Assessment of blue-green algal toxins in raw and finished drinking water
(No. 90815). AWWA Research Foundation, Denver, CO.
Carstea, E.M., Baker, A., Bieroza, M., Reynolds, D.M., Bridgeman, J., 2014. Characterisation of
dissolved organic matter fluorescence properties by PARAFAC analysis and thermal
quenching. Water Res. 61:152-161.
Chang, D.-W., Hobson, P., Burch, M., Lin, T.-F., 2012. Measurement of cyanobacteria using in-
vivo fluoroscopy - Effect of cyanobacterial species, pigments, and colonies. Water Res.
46(16):5037-5048.
Che, H., Liu, S., 2014. Contaminant detection using multiple conventional water quality sensors
in an early warning system. Procedia Eng. 89:479-487.
Chen, B., Wu, H., Li, S.F.Y., 2014. Development of variable pathlength UV-vis spectroscopy
combined with partial-least-squares regression for wastewater chemical oxygen demand
(COD) monitoring. Talanta 120:325-330.
Chen, J., Chen, H., Geng, H., Li, H., Qi, J., 2010. Development of a GIS-based oil spill model
and its application in the Three Gorges Reservoir. J. Environ. Hydrol. 18(17).
CNEMC, 2009. National surface water quality automatic monitoring system introduced [WWW
Document], URL http://www.cnemc.cn/publish/totalWebSite/news/news 5027.html
(accessed February 6, 2016).
Cracknell, A.P., Newcombe, S.K., Black, A.F., Kirby, N.E., 2001. The ABDMAP (Algal Bloom
Detection, Monitoring and Prediction) concerted action. Int. J. Remote Sens. 22(2-
3):205-247.
Dawsey, W.J., Minsker, B.S., VanBlaricum, V.L., 2006. Bayesian belief networks to integrate
monitoring evidence of water distribution system contamination. J. Water Resour. Plan.
Manag. 132(4):234-241.
Dehua, W., Pan, L., Bo, L., Zeng, G., 2012. Water quality automatic monitoring system based on
GPRS data communications. 2012 Int. Conf. Mod. Hydraul. Eng., Procedia Eng. 28:840-
843.
Delpla, I., Jung, A.-V., Baures, E., Clement, M., Thomas, O., 2009. Impacts of climate change
on surface water quality in relation to drinking water production. Environ. Int.
35(8): 1225—1233.
Deng, Z., 2014. Development of water pollution early warning system for oyster harvesting
areas. 11th International Conference on Hydroinformatics, NY, U.S.
Dolman, A.M., Riicker, J., Pick, F.R., Fastner, J., Rohrlack, T., Mischke, U., Wiedner, C., 2012.
Cyanobacteria and cyanotoxins: the influence of nitrogen versus phosphorus. PloS One
7(6):e38757.
Downing, B.D., Pellerin, B.A., Bergamaschi, B.A., Saraceno, J.F., Kraus, T.E.C., 2012. Seeing
the light: The effects of particles, dissolved materials, and temperature on in situ
measurements of DOM fluorescence in rivers and streams. Limnol. Oceanogr. Methods
10:767-775.
DRBC, 2013. Basin information. Delaware River Basin Commission. Retrieved January 18,
2016,	from http://www.ni .gov/drbc/basin.
Drolc, A., Vrtovsek, J., 2010. Nitrate and nitrite nitrogen determination in waste water using on-
line UV spectrometric method. Bioresour. Technol. 101(11):4228-4233.
Duzinski, P., 2008. Delaware Valley early warning system (EWS) - Regional Response Team
III, Pittsburgh, PA. September 24, 2008. [WWW Document], Retrieved September 19,
2017,	from https://www.slidedoc.us/delaware-vallev-early-warning-svstem-ews
57

-------
Eltzov, E., Marks, R.S., Voost, S., Wullings, B.A., Heringa, MB., 2009. Flow-through real time
bacterial biosensor for toxic compounds in water. Sens. Actuators B Chem. 142(1): 11-
18.
Etheridge, JR., Birgand, F., Burhell, M.R., II., Smith, B.T., 2013. Addressing the fouling of in
situ ultraviolet-visual spectrometers used to continuously monitor water quality in
brackish tidal marsh waters. J. Environ. Qual. 42(6): 1896-1901.
European Commission, 2012. EU - China River Basin Management Programme [WWW
Document], Retrieved September 19, 2017, from
http://eeas.europa.eu/archives/delegations/china/documents/eu china/development coope
ration/rbmp 6083 20120627.pdf
Fennell, S., 1988. Ohio River Model for contaminant plume time-of-arrival and peak
concentration estimation. ORSANCO, Cincinnati, OH.
Francy, D.S., Graham, J.L., Stelzer, E.A., Ecker, C.D., Brady, A.M.G., Struffolina, P., Loftin,
K.A., 2015. Water quality, cyanobacteria, and environmental factors and their relations to
microcystin concentrations for use in predictive models at Ohio Lake Erie and Inland
Lake Recreational Sites, 2013-14: U.S. Geological Survey [USGS] Scientific
Investigations Report 2015-5120. USGS, Reston, VA.
Francy, D., Stelzer, E., 2014. Molecular methods for detecting and quantifying cyanobacteria.
U.S. EPA Inland HAB Webinar January 14, 2014. URL
https://www.epa.gov/sites/production/files/2014-
01/documents/molecular methods for detecting and quantifying cyanobacteria francy
and stelzer usgs.pdf (accessed on January 19, 2016).
Freeman, K.S., 2011. Forecasts aid HABs [harmful algal blooms] response. Environ. Health
Perspect. 119:A510-A510.
Gehringer, M.M., Wannicke, N., 2014. Climate change and regulation of hepatotoxin production
in cyanobacteria. FEMS Microbiol. Ecol. 88(1): 1-25.
Girotti, S., Ferri, E.N., Fumo, M.G., Maiolini, E., 2008. Monitoring of environmental pollutants
by bioluminescent bacteria. Anal. Chim. Acta 608(l):2-29.
Glasgow, H.B., Burkholder, J.M., Reed, R.E., Lewitus, A.J., Kleinman, J.E., 2004. Real-time
remote monitoring of water quality: a review of current applications, and advancements
in sensor, telemetry, and computing technologies. J. Exp. Mar. Biol. Ecol. 300(l-2):409-
448.
Graham, J.L., Loftin, K.A., Kamman, N., 2009. Monitoring recreational freshwaters. LakeLine,
Summer 2009, North American Lake Management Society, Madison, WI.
Graham, J.L., Loftin, K.A., Meyer, M.T., Ziegler, A.C., 2010. Cyanotoxin mixtures and taste-
and-odor compounds in cyanobacterial blooms from the midwestern United States.
Environ. Sci. Technol. 44(19):7361-7368.
Grayman, W. M., 2008. Designing an optimum water monitoring system. Wiley Handbook of
Science and Technology for Homeland Security. 3(2): 1-33.
Grayman, W.M., Deininger, R.A., Males, R.M., 2001. Design of early warning and predictive
source-water monitoring systems (No. 90878), Monitoring and Analysis. AWWA
Research Foundation, Denver, CO.
Grayman, W.M., Kshirsagar, S.R., Males, R.M., Goodrich, J. A., Heath, J.P., 1994. A GIS for the
Ohio River basin. Presented at the National Conference on Environmental Problem-
Solving with Geographic Information Systems, Cincinnati, OH.
Grayman, W.M., Males, R.M., 2002. Risk-based modeling of early warning systems for
58

-------
pollution accidents. Wat. Sci. Tech., 46(3):29-41.
Gullick, R.W., 2003. Monitoring systems for early warning of source water contamination.
American Water Works Service Company, Inc., Voorhees, NJ.
Gullick, R.W., Gaffney, L.J., Crockett, C.S., Schulte, J., Gavin, A.J., 2004. Developing regional
early warning systems for U.S. source waters. J. Am. Water Works Assoc. 96(6):68-82.
Gullick, R.W., Grayman, W., M., Deininger, R.A., Males, R.M., 2003. Design of early warning
monitoring systems for source waters. J. Am. Water Works Assoc. 5:58-72.
Gunatilaka, A., Moscetta, P., Sanfilippo, L., 2007. Recent advancements in water quality
monitoring - the use of miniaturized sensors and novel analytical measuring techniques
for in-situ and on-line real time measurements. In: SYSTEA -Systems Technology
Advance SpA, 2A - 03012. Presented at the International Workshop on Monitoring and
Sensor for Water Pollution Control, Beijing, China.
Guo, W., Xu, J., Wang, J., Wen, Y., Zhuo, J., Yan, Y., 2010. Characterization of dissolved
organic matter in urban sewage using excitation emission matrix fluorescence
spectroscopy and parallel factor analysis. J. Environ. Sci. 22(11): 1728-1734.
Hadeed, S.J., 1978. Early organic detection system established on the Ohio River. J. Water.
Pollut. Con. 50(3):416-418.
Hall, J., Szabo, J.G., 2010. Online water quality monitoring in drinking water distribution
systems: A summary of U.S. EPA research and best practices. J. Am. Water Works
Assoc. 102(8), 20-22.
Hall, J.S., Szabo, J.G., Panguluri, S., Meiners, G, 2009. Distribution System Water Quality
Monitoring: Sensor Technology Evaluation Methodology and Results (EPA 600/R-
09/076). U.S. EPA, Office of Research and Development, Cincinnati, OH.
Han, S., Zhao, Y., Yu, T., Craik, S., 2014. Challenge studies of the Hach GuardianBlue early
warning system. J. Environ. Eng. 140:A4014002.
Hart, D.B., McKenna, S.A., Murray, R., Haxton, T., 2011. Combining water quality and
operational data for improved event detection. Presented at the Water Distribution
Systems Analysis 2010, ASCE, Tuscon, AZ.
Hart, W.E., Murray, R., 2010, Review of sensor placement strategies for contamination warning
systems in drinking water distribution systems. J. Water Resour. Plan. Manag.
13 6(6): 611-619.
Henderson, R.K., Baker, A., Murphy, K.R., Hambly, A., Stuetz, R.M., Khan, S.J., 2009.
Fluorescence as a potential monitoring tool for recycled water systems: A review. Water
Res. 43(4):863-881.
Horn, C.R., Grayman, W.M., 1993. Water-quality modeling with EPA Reach File system. J.
Water Resour. Plan. Manag. 119(2): 262-274. doi:10.1061/(ASCE)0733-
9496(1993)119:2(262).
Howard, B., 2007. Real Time Monitoring Program - Protecting the drinking water source in the
Huron to Erie Corridor [WWW Document], Retrieved September 20, 2017 from
http://www.michigan.gov/documents/deq/deq-wb-wws-BrockHowardCIPRTMtalk5-30-
07 237078 7.pdf
Hu, Y., Wen, Y., Wang, X., 2016. Novel method of turbidity compensation for chemical oxygen
demand measurements by using UV-vis spectrometry. Sens. Actuators B Chem.
277:393-398.
Hunter, P.D., Tyler, A.N., Willby, N.J., Gilvear, D.J., 2008. The spatial dynamics of vertical
migration by Microcystis aeruginosa in a eutrophic shallow lake: A case study using high
59

-------
spatial resolution time-series airborne remote sensing. Limnol. Oceanogr. 53(6):2391-
2406
ICPDR, 2016. AEWS - Accident Emergency Warning System [WWW Document], Retrieved on
September 20, 2017 from https://www.icpdr.org/main/activities-proiects/aews-accident-
emergency-warning-system
ICPDR, 2015a. The Danube River Basin District Management Plan, Part A - Basin-wide
overview [WWW Document], URL https://www.icpdr.org/main/management-plans-
danube-river-basin-published (accessed January 21, 2016)
ICPDR, 2015b. Transnational Monitoring Network - Surface Waters [WWW Document], URL
http://www.icpdr.org/main/publications/maps-danube-river-basin-district-management-
plan-2015 (accessed February 1, 2016)
ICPDR, 2014. International Operations Manual for PIACs of the Danube AEWS [WWW
Document], URL https://www.icpdr.org/main/activities-proiects/aews-accident-
emergencv-warning-svstem (accessed February 1, 2016).
ICPDR, 2011. Danube Basin Facts and Figures [WWW Document], URL
https://www.icpdr.org/main/danube-basin/countries-danube-river-basin (accessed
February 1, 2016).
IWAC (International Water Assessment Centre), 2001. Assessment Practices and Environmental
Status of Ten Transboundary Rivers in Europe [WWW Document], URL
http://www.iwacportal.org/File/downloads/10 rivers report.pdf (accessed February 1, 2016).
Izydorczyk, K., Carpentier, C., Mrowczynski, J., Wagenvoort, A., Jurczak, T., Tarczynska, M.,
2009. Establishment of an alert level framework for cyanobacteria in drinking water
resources by using the algae online analyser for monitoring cyanobacterial chlorophyll a.
Water Res. 43(4):989-996.
Izydorczyk, K., Tarczynska, M., Jurczak, T., Mrowczynski, J., Zalewski, M., 2005.
Measurement of phycocyanin fluorescence as an online early warning system for
cyanobacteria in reservoir intake water. Environ. Toxicol. 20(4):425-430.
Jeong, J.-J., Bartsch, S., Fleckenstein, J.H., Matzner, E., Tenhusen, J.D., Lee, S.D., Park, S.K.,
Park, J.-H., 2012. Differential storm responses of dissolved and particulate organic
carbon in a mountainous headwater stream, investigated by high-frequency, in situ
optical measurements. J. Geophys. Res. 117(G03013).
Jiang, P., Xia, H., He, Z., Wang, Z., 2009. Design of a water environment monitoring system
based on wireless sensor networks. Sensors 9:6411-6434.
Johnstone, D.W., Sanchez, N.P., Miller, C.M., 2009. Parallel factor analysis of excitation-
emission matrices to assess disinfection byproduct formation during a peak formation
period. Environ. Eng. Sci. 26(10): 1551—1559.
Khan, A., Khan, H., 2008. An environmental security and water resources management system
using real time water quality warning and communication for the Nile River. Presented at
the World Environmental and Water Resources Congress, ASCE, Honolulu, Hawaii.
Khan, H., Dawe, P., Paterson, R., Abdel-Gawad, S., 2011. Water resources management system
for the Nile River. Presented at the International Conference on Environment Science and
Engineering, CBEES, IEEE, Bali Island, Indonesia.
Klemas, V., 2012. Remote sensing of algal blooms: An overview with case studies. J. Coast.
Res. 28(lA):34-43.
Kokkali, V., van Delft, W., 2014. Overview of commercially available bioassays for assessing
chemical toxicity in aqueous samples. TrAC Trends Anal. Chem. 61:133-155.
60

-------
Korshin, G.V., Li, C.-W., Benjamin, M.M., 1997. Monitoring the properties of natural organic
matter through UV spectroscopy: A consistent theory. Water Res. 31(7): 1787-179.
Kudela, R.M., Palacios, S.L., Austerberry, D.C., Accorsi, E.K., Guild, L.S., Torres-Perez, J.,
2015. Application of hyperspectral remote sensing to cyanobacterial blooms in inland
waters. Remote Sens. Environ., Special Issue on the Hyperspectral Infrared Imager
(HyspIRI) 167:196-205.
Kutser, T., Metsamaa, L., Strombeck, N., Vahtmae, E., 2006. Monitoring cyanobacterial blooms
by satellite remote sensing. Estuar. Coast. Shelf Sci. 67(l-2):303-312.
Langergraber, G., Fleischmann, N., Hofstadter, F., 2003. A multivariate calibration procedure for
UV/VIS spectrometric quantification of organic matter and nitrate in wastewater. Water
Sci. Technol. 47(2):63-71.
Lee, J.H.W., Hodgkiss, I.J., Wong, K.T.M., Lam, I.H.Y., 2005. Real time observations of coastal
algal blooms by an early warning system. Estuar. Coast. Shelf Sci. 65(1-2): 172-190.
Li, W.-T., Jin, J., Li, Q., Wu, C.-F., Lu, H., Zhou, Q., Li, A.-M., 2016. Developing LED UV
fluorescence sensors for online monitoring DOM and predicting DBPs formation
potential during water treatment. Water Res. 93:1-9.
Lichota, S., DeMaria, A., 2009. St. Clair River - Lake St. Clair - Detroit River, Drinking Water
Protection Network [WWW Document], URL
https://acwi.gov/monitoring/conference/2008/presentations/sessionJ/J4C_DeMaria.pdf
(accessed January 21, 2016).
Liu, S., Smith, K., Che, H., 2015. A multivariate based event detection method and performance
comparison with two baseline methods. Water Res. 80:109-118.
Loftin, K.A., Graham, J.L., Hilborn, E.D., Lehmann, S.C., Meyer, M.T., Dietz, J.E., Griffith,
C.B., 2008. Cyanotoxins in inland lakes of the United States: Occurrence and potential
recreational health risks in the EPA National Lakes Assessment 2007. Harmful Algae
56:77-90.
Louisiana DEQ, 2016. Early Warning Organic Compound Detection System. Retrieved January
30, 2016, from
http://wwwl.deq.louisiana.gov/portal/DIVISIONS/Surveillance/EarlvWarningOrganicCo
mpoundDetectionSvstem.aspx
Louisiana DEQ, 2014. Early Warning System helps protect the Mississippi River as a drinking
source. Discover DEQ, Louisiana Department of Environmental Quality Newsletter,
November 2014, Issue Number: 34.
Louisiana DEQ, 2012. Retrieved January 31, 2016, from
http://map.ldeq.org/images/maps/201201036.jpg.
Louisiana DEQ, 2009. Commission on Streamlining Government Executive Summary. Retrieved
September 25, 2017, from
http://wwwl.deq.louisiana.gov/portal/Portal s/0/news/DEQ%20Executive%20Summarv%
20Report%20Commission%20on%20Streamlining%20Government%20081709.pdf
Louren<;o, N.D., Chaves, C.L., Novais, J.M., Menezes, J.C., Pinheiro, H.M., Diniz, D., 2006. UV
spectra analysis for water quality monitoring in a fuel park wastewater treatment plant.
Chemosphere 65(5):786-79.
Loza, V., Perona, E., Mateo, P., 2013. Molecular fingerprinting of cyanobacteria from river
biofilms as a water quality monitoring tool. Appl. Environ. Microbiol. 79(5): 1459-1472.
Lunetta, R.S., Schaeffer, B.A., Stumpf, R.P., Keith, D., Jacobs, S.A., Murphy, M.S., 2015.
Evaluation of cyanobacteria cell count detection derived from MERIS imagery across the
61

-------
eastern U.S.A. Remote Sens. Environ., Special Issue: Remote Sensing of Inland Waters
157:24-34.
Maier, H.R., Jain, A., Dandy, G.C., Sudheer, K.P., 2010. Methods used for the development of
neural networks for the prediction of water resource variables in river systems: Current
status and future directions. Environ. Model. Softw. 25(8):891-909.
Maradona, A., Marshall, G., Mehrvar, M., Pushchak, R., Laursen, A.E., McCarthy, L.H., Bostan,
V., Gilbride, K.A., 2012. Utilization of multiple organisms in a proposed early-warning
biomonitoring system for real-time detection of contaminants: Preliminary results and
modeling. J. Hazard. Mater. 219-220:95-102.
Matthews, M.W., Bernard, S., Winter, K., 2010. Remote sensing of cyanobacteria-dominant
algal blooms and water quality parameters in Zeekoevlei, a small hypertrophic lake, using
MERIS. Remote Sens. Environ. 114(9):2070-2087.
McHarg, I.L., 1969. Design with Nature. Natural History Press, Garden City, N.Y.
McKenna, S.A., Wilson, M., Klise, K.A., 2008. Detecting changes in water quality data. J. Am.
Water Works Assoc. 100(l):74-85.
McQuaid, N., Zamyadi, A., Prevost, M., Bird, D.F., Dorner, S., 2011. Use of in vivo
phycocyanin fluorescence to monitor potential microcystin-producing cyanobacterial
biovolume in a drinking water source. J. Environ. Monit. 13:455-463.
Ministry of Environmental Protection, 2015. MEP [Ministry of Environmental Protection]
announces the state of the environment in China in the first six months, July 25, 2015
[WWW Document], URL http://english.mep.gov.cn/News service/
news_release/201508/t20150817_308198.htm (accessed February 6, 2016)
Moreira, C., Ramos, V., Azevedo, J., Vasconcelos, V., 2014. Methods to detect cyanobacteria
and their toxins in the environment. Appl. Microbiol. Biotechnol. 98(19):8073-8082.
Murray, R., Haxton, T., Janke, R., Hart, W.E., Berry, J., Phillips, C., 2010. Sensor network
design for drinking water contamination warning systems: a compendium of research
results and case studies using the TEVA-SPOT software. (EPA 600/R-09/0141B). U.S.
EPA, Office of Research and Development, Cincinnati, OH.
Murray, S., Ghazali, M., McBean, E.A., 2011. Real-time water quality monitoring: Assessment
of multisensor data using Bayesian belief networks. J. Water Resour. Plan. Manag.
138:63-70.
NexSens Technology, 2016. St. Clair & Detroit River Monitoring, Michigan Department of
Environmental Quality. NexSens Technology, Inc., Fairborn, OH. [WWW Document],
URL http://www.nexsens.com/case_studies/st-clair-detroit-river-monitoring.htm
(accessed January 21, 2016).
NOAA, 2013. NOAA Harmful Algal Bloom Operational Forecast System (HAB-OFS). [WWW
Document], URL https://tidesandcurrents.noaa.gov/hab/
NOAA GLERL, 2016. Great Lakes harmful algal blooms (HABs) and hypoxia. NOAA Great
Lakes Environmental Research Laboratory (GLERL) Great Lakes Water Quality [WWW
Document], Retrieved September 20, 2016 from
https://www.glerl.noaa.gov/res/HABs and Hypoxia/
NOAA GLERL Research Laboratory, 2015a. Experimental Lake Erie Harmful Algal Bloom
Bulletin, Bulletin, 27. Ann Arbor, MI: NOAA Great Lakes Environmental Research
Laboratory.
NOAA GLERL, 2015b. Experimental Lake Erie Harmful Algal Bloom Bulletin, Bulletin 17.
Ann Arbor, MI: NOAA Great Lakes Environmental Research Laboratory.
62

-------
NOAA GLERL, 2015c. Experimental Lake Erie Harmful Algal Bloom Bulletin, Bulletin 16.
Ann Arbor, MI: NOAA Great Lakes Environmental Research Laboratory.
Niibel, U., Garcia-Pichel, F., Muyzer, G., 1997. PCR primers to amplify 16S rRNA genes from
cyanobacteria. Appl. Environ. Microbiol. 63(8):3327-3332.
O'Halloran, R., Fogelman, S., Zhao, H., 2009. Current Online Water Quality Monitoring
Methods and their Suitability for the Western Corridor Purified Recycled Water Scheme
(Urban Water Security Research Alliance Technical Report, No. 10). The Urban Water
Security Research Alliance, Queensland, Australia.
Ohio Environmental Protection Agency, 2014. Draft Public Water System Harmful Algal Bloom
Response Strategy. [WWW Document], Retrieved September 25 from
http: //epa. ohi o. gov/ddagw/HAB. aspx
Oliker, N., Ostfeld, A., 2015. Network hydraulics inclusion in water quality event detection
using multiple sensor stations data. Water Res. 80:47-58.
Oliker, N., Ostfeld, A., 2014. Minimum volume ellipsoid classification model for contamination
event detection in water distribution systems. Environ. Model. Softw. 57:1-12.
O'Reilly, C.M., Sharma, S., Gray, D.K., Hampton, S.E., Read, J.S., Rowley, R.J., Schneider, P.,
Lenters, J.D., Mclntyre, P.B., Kraemer, B.M., Weyhenmeyer, G.A., Straile, D., Dong, B.,
Adrian, R., Allan, M.G., Anneville, O., Arvola, L., Austin, J., Bailey, J.L., Baron, J.S.,
Brookes, J.D., de Eyto, E., Dokulil, M.T., Hamilton, D.P., Havens, K., Hetherington,
A.L., Higgins, S.N., Hook, S., Izmest'eva, L.R., Joehnk, K.D., Kangur, K., Kasprzak, P.,
Kumagai, M., Kuusisto, E., Leshkevich, G., Livingstone, D.M., Maclntyre, S., May, L.,
Melack, J.M., Mueller-Navarra, D.C., Naumenko, M., Noges, P., Noges, T., North, R.P.,
Plisnier, P.-D., Rigosi, A., Rimmer, A., Rogora, M., Rudstam, L.G., Rusak, J.A.,
Salmaso, N., Samal, N.R., Schindler, D.E., Schladow, S.G., Schmid, M., Schmidt, S.R.,
Silow, E., Soylu, M.E., Teubner, K., Verburg, P., Voutilainen, A., Watkinson, A.,
Williamson, C.E., Zhang, G., 2015. Rapid and highly variable warming of lake surface
waters around the globe. Geophys. Res. Lett. 42(24):10,773-10,781.
ORSANCO, 2016a. Ohio River Basin. [WWW Document], Ohio River Valley Water Sanitation
Commission, Cincinnati, OH. URL http://www.orsanco.org/ohio-river-basin (accessed
February 7, 2016)
ORSANCO, 2016b. Organics Detection System [WWW Document], Ohio River Valley Water
Sanitation Commission, Cincinnati, OH. URL http://www.orsanco.org/ organics-
detection-system-84 (accessed February 7, 2016)
Ouyang, H.-T., Yu, H., Lu, C.-H., Luo, Y.-H., 2008. Design optimization of river sampling
network using genetic algorithms. J. Water Resour. Plan. Manag. 134:83-87.
Paerl, H.W., Huisman, J., 2009. Climate change: A catalyst for global expansion of harmful
cyanobacterial blooms. Environ. Microbiol. Rep. 1(1):27—37.
Paerl, H.W., Paul, V.J., 2012. Climate change: Links to global expansion of harmful
cyanobacteria. Water Res. 46(5): 1349-1363.
Panguluri, S., Grayman, W., Clark, R., 2005. Water distribution system analysis: Field studies,
modeling and management [A reference guide for utilities] (EPA/600/R-06/028). U. S.
Environmental Protection Agency, Cincinnati, Ohio.
Park, C., Kim, S.-H., Telci, I.T., Aral, M.M., 2010. Designing optimal water quality monitoring
network for river systems and application to a hypothetical river, in: Proceedings of the
63

-------
Winter Simulation Conference, WSC 2010. Winter Simulation Conference, Baltimore,
Maryland, pp. 3506-3513.
Park, S.-Y., Choi, J.H., Wang, S., Park, S.S., 2006. Design of a water quality monitoring network
in a large river system using the genetic algorithm. Ecol. Model. 199(3):289-297.
Perelman, L., Arad, J., Housh, M., Ostfeld, A., 2012. Event detection in water distribution
systems from multivariate water quality time series. Environ. Sci. Technol. 46(15):8212-
8219.
Piotrowski, A., Wallis, S.G., Napiorkowski, J.J., Rowinski, P.M., 2007. Evaluation of 1-D tracer
concentration profile in a small river by means of multi-layer perceptron neural networks.
Hydrol Earth Syst Sci 11:1883-1896.
Rea, A.H., Clark, E.P., Adams, A., Samuels, W.B., 2015. The Open Water Data Initiative: Water
information for a thirsty nation. Water Resour. Impact 17(6):7-10.
River Alert Information Netork, 2016. [WWW Document], URL: http://www.rainmatters.org/.
Rain River Alert Information Network (accessed December 6, 2016).
Reiger, L., Langergraber, G., Thomann, M., Fleischmann, N., Siegrist, H., 2004. Spectral in-situ
analysis of NO2, NO3, COD, DOC and TSS in the effluent of a WWTP. Water Sci.
Technol. 50:143-152.
Ren, Z., Wang, Z., 2010. Differences in the behavior characteristics between Daphnia magna
and Japanese madaka in an on-line biomonitoring system. J. Environ. Sci. 22(5):703-708.
Richardson, T.L., Lawrenz, E., Pinckney, J.L., Guajardo, R.C., Walker, E.A., Paerl, H.W.,
Maclntyre, H.L., 2010. Spectral fluorometric characterization of phytoplankton
community composition using the Algae Online Analyserฎ. Water Res. 44(8):2461-
2472.
Roccaro, P., Yan, M., Korshin, G.V., 2015. Use of log-transformed absorbance spectra for online
monitoring of the reactivity of natural organic matter. Water Res. 84:136-143.
Rosen, J.S., Kearns, M., 2015. Technologies/innovation to optimize water quality management.
Presented at the Water Quality Technology Conference, American Water Works
Association, Salt Lake City, UT.
Rosen, J.S., Whelton, A.J., McGuire, M.J., Clancy, J.L., Bartrand, T.A., Eaton, A., Patterson, J.,
Dourson, M., Nance, P., Adams, C., 2014. The crude MCHM chemical spill in
Charleston, W.Va. J. AWWA 106(9):65-74.
Rui, Y., Shen, D., Khalid, S., Yang, Z., Wang, J., 2015. GIS-based emergency response system
for sudden water pollution accidents. Phys. Chem. Earth Parts ABC, Water Scarcity and
Integrated Water Resources Management 79-82:115-121.
Samuels, W.B., Amstutz, D.E., Bahadur, R., Pickus, J.M., 2006. RiverSpill: A national
application for drinking water protection. J. Hydraul. Eng. 132(4):393-403.
Samuels, W.B., Bahadur, R., Ziemniak, C., Amstutz, D.E., 2015. Development and application
of the incident command tool for drinking water protection. Water Environ. J. 29(1): 1-
15.
Sanchez, N.P., Skeriotis, A.T., Miller, C.M., 2014. A PARAFAC-based long-term assessment of
DOM in a multi-coagulant drinking water treatment scheme. Environ. Sci. Technol.
48(3): 1582—1591.
Schulte, J.G., 2014. Early Warning Detection Training Workshop. Morgantown, WV.
http://www.horslevwitten.com/sourcewater/pdf/Schulte.pdf (accessed January 20, 2016).
Simis, S.G.H., Peters, S.W.M., Gons, H.J., 2005. Remote sensing of the cyanobacterial pigment
phycocyanin in turbid inland water. Limnol. Oceanogr. 50(l):237-245.
64

-------
Solomatine, D., Ostfeld, A., 2008. Data-driven modelling: Some past experiences and new
approaches. J. Hydroinformatics 10(l):3-22.
SourcewaterPA, 2015. Early Warning Systems: A Key SWP Tool [WWW Document], URL
http://www.sourcewaterpa.ore/7page id=l806 (accessed January 2, 2016)
Spreafico, M., van Mazijk, A., 1993. Alarm Model Rhine: A Model for the Operational
Prediction of the Transport of Pollutants in the Rhine River (No. 1-12 of the CHR).
IRC/CHR Committee of Experts, Lelystad, Germany.
Spruce, J.P., 2004. Decision Support Tool Evaluation Report for General NOAA Oil Modeling
Environment (GNOME) Version 2.0. NASA, John C. Stennis Space Center, MS.
SRBC, 2016. Early Warning System - Overview. Susquehanna River Basin Commission,
Harrisburg, PA. [WWW Document], URL http://www.srbc.net/drinkingwater/ (accessed
September 25, 2017).
SRBC, 2015. Remote Water Quality Monitoring Network, Data Report of Baseline Conditions
for 2010 - 2013 [WWW Document], URL
http://mdw.srbc.net/remotewaterqualitv/assets/downloads/
pdf/RWQMN_datareport_2010-2013.PDF (accessed January 19, 2016)
SRBC, 2013a. Information Sheet: Susquehanna River Basic Facts [WWW Document], URL
https://www.srbc.net/pubinfo/docs/infosheets/Susq%20River%20Basin%20Facts Fact%
20Sheet FINAL2017.PDF (accessed September 25, 2017).
SRBC, 2013b. State of the Susquehanna, 2013 Report [WWW Document], URL
http://www.srbc.net/stateofsusq2013/docs/2013 SOTS Report Final low res.pdf
(accessed September 21, 2017).
SRBC, 2012. Information Sheet: Susquehanna River Basin Early Warning System [WWW
Document], URL http://www.srbc.net/programs/docs/EWSInfoSheet020712.PDF
(accessed January 19, 2016).
Srivastava, A., Singh, S., Ahn, C.-Y., Oh, H.-M., Asthana, R.K., 2013. Monitoring approaches
for atoxic cyanobacterial bloom. Environ. Sci. Technol. 47(16): 8999-9013.
doi:10.1021/es401245k
Stedmon, C.A., Seredynska-Sobecka, B., Boe-Hansen, R., Le Tallec, N., Waul, C., Arvin, E.,
2011. A potential approach for monitoring drinking water quality from groundwater
systems using organic matter fluorescence as an early warning for contamination events.
Water Res 45(18):6030-6038.
Storey, M.V., van der Gaag, B., Burns, B.P., 2011. Advances in on-line drinking water quality
monitoring and early warning systems. Water Res. 45(2):741-747.
Streeter, H.W., Phelps, E.B., 1925. A study of the pollution and natural purification of the Ohio
river. III. Factors concerned in the phenomena of oxidation and reaeration. Public Health
Bulletin no. 146, Reprinted by U.S. Department of Health, Education and Welfare, Public
Health Service, 1958.
Strobl, R.O., Robillard, P.D., 2008. Network design for water quality monitoring of surface
freshwaters: A review. J. Environ. Manage., Microbial and Nutrient Contaminants of
Fresh and Coastal Waters 87(4):639-648.
Stumpf, R.P., Wynne, T.T., Baker, D.B., Fahnenstiel, G.L., 2012. Interannual variability of
cyanobacterial blooms in Lake Erie. PLoS ONE 7(8):e42444
Swanson, G., 2012. A unique source water monitoring initiative on the Upper Mississippi River.
Watercon, 2012, Illinois Water Environment Association., Springfield, IL. Retrieved
September 25, 2017, from
65

-------
https://c.vmcdn.com/sites/www.isawwa.org/resource/resmgr/watercon2012-tuesdav-
pdf/tuerespot830.pdf
Telci, I.T., Nam, K., Guan, J., Aral, M.M., 2009. Optimal water quality monitoring network
design for river systems. J. Environ. Manage. 90:2987-2998.
Thomann, R.V., 1963. Mathematical model for dissolved oxygen. ASCE J. Sanit. Eng. 89:1-32.
Tomlinson RF., 1968. A geographic information system for regional planning. In Stewart, G.A.
(ed): Symposium on Land Evaluation, Commonwealth Scientific and Industrial Research
Organization, MacMillan of Australia., Melbourne, 1968.
Trescott, A., 2012. Remote Sensing Models of Algal Blooms and Cyanobacteria in Lake
Champlain. Masters thesis, University of Massachusetts, Amherst.
UMRBA, 2016. Upper Mississippi River Basin Association. UMBRA, St. Paul, MN. Retrieved
January 19, 2016, from http://www.umrba.org/index.htm
UMRBA, 2014. Upper Mississippi River Spill Response Plan & Resource Manual. UMBRA, St.
Paul, MN. Retrieved January 20, 2016, from http://www.umrba.org/hazspills/umrplan.pdf
UMRBA, 2007. Pilot Monitoring Station at Lock & Dam 15: Project Evaluation Report.
UMBRA, St. Paul, MN. Retrieved January 19, 2016, from
http://www.umrba.org/publications/hazspills/piloteval.pdf
UNEP, 2012. Early Warning Systems: A State of the Art Analysis and Future Directions.
Division of Early Warning and Assessment (DEW A), United Nations Environment
Programme (UNEP), Nairobi.
USACE, 2016. HEC-RAS [Hydraulic Engineering Center-River Analysis System] Features
[WWW Document], URL http://www.hec.usace.army.mil/software/hec-ras/features.aspx
(accessed January 14, 2016).
U.S. EPA, 2016a. DWMAPS - Drinking Water Mapping Application [WWW Document], URL
https://geopub.epa.gov/DWWidgetApp/ and
https://www.epa.gov/sourcewaterprotection/dwmaps.
U.S. EPA, 2016b. Guidance for Designing Communications Systems: For Water Quality
Surveillance and Response Systems (EPA 817-B-16-002). U.S. EPA, Office of Water,
Washington DC.
U.S. EPA, 2016c. Online Source Water Quality Monitoring: For Water Quality Surveillance and
Response Systems (EPA 817-B-16-003). U.S. EPA, Office of Water, Washington DC.
U.S. EPA, 2015. Water Quality Surveillance and Response System Primer. (EPA 817-B-15-
002). U.S. EPA, Office of Water, Washington DC.
U.S. EPA, 2012. User's Manual for CANARY - Version 4.3.2 (EPA 600/R-08/040B). U.S. EPA,
Office of Research and Development, Cincinnati, OH.
U.S. EPA, 2005. Technologies and Techniques for Early Warning Systems to Monitor and
Evaluate Drinking Water Quality: A State-of-the-Art Review (EPA/600/R-05/156). U.S.
EPA, Office of Water, Washington DC.
U.S. EPA, 2002. Guidance for Quality Assurance Project Plans for Modeling EPA QA/G-5M
(Policies and Guidance EPA/240/R-02/007), U.S. EPA, Office of Environmental
Information, Washington DC.
Vaillant, S., Pouet, M.F., Thomas, O., 2002. Basic handling of UV spectra for urban water
quality monitoring. Urban Water 4(3):273-281.
van den Broeke, J., Carpentier, C., Moore, C., Carsen, L., Jonsson, J., Sivil, D., Rosen, J.S.,
Cade, L., Mofidi, A., Swartz, C., Coomans, N., 2014. Compendium of sensor and
66

-------
monitors and their use in the global water industry (No. SENG1C11). WERF,
Alexandria, VA.
van den Broeke, J., Ross, P.S., van der Helm, A.W., Baars, E.T., Rietveld, L.C., 2008. Use of on-
line UV/Vis-spectrometry in the measurement of dissolved ozone and AOC
concentrations in drinking water treatment. Water Sci. Technol. 57(8): 1169-1175.
VRIC, EI, 2014. Water Quality Early Warning System - On transboundary watercourses of the
Tisza River Basin, Chapters 1-4. [WWW Document], Veszprem, Hungary, Vespremi
Regionalis Innovacios Centrum (Regional Center of Innovation). URL
https://www.danubewaterquality.eu/uploads/mod_files/WQM-EWS_part-l-
4_EN_v2.0_.pdf (accessed September 25, 2017).
Waldon, M.G., 1998. Time-of-travel in the Lower Mississippi River: Model development,
calibration, and application. Water Environ. Res. 70(6): 1132-1141.
Wang, Q., Li, S., Jia, P., Qi, C., Ding, F., Wang, Q., Li, S., Jia, P., Qi, C., Ding, F., 2013. A
review of surface water quality models. Sci. World J. 2013 :e231768.
Wilhelm, S.W., Ouellette, A. J. A., Rinta-Kanto, J., 2007. Development of Molecular Reporters
for Microcystis Activity and Toxicity (Project No. 2818). Awwa Research Foundation,
Denver, CO.
Wold, A., 2015. Pollution monitoring program gets boost thanks to money from a settlement last
year between ExxonMobil and Louisiana. August 25, 2015, The Advocate.
Woutersen, M., Belkin, S., Brouwer, B., van Wezel, A.P., Heringa, M.B., 2011. Are luminescent
bacteria suitable for online detection and monitoring of toxic compounds in drinking
water and its sources? Anal. Bioanal. Chem. 400(4):915-929.
Wrubel, B., 2014. Underwater Detective: The Importance of Near-Real Times River Water
Monitoring. Slide presentation. City of Marysville, MI. [WWW Document], URL http://
www.friendsofstclair.ca/www/pdf/2014%20Symposium/Drinking%20Water%20In take
%20Monitoring%20-%20Wrubel.pdf (accessed September 25, 2017).
Wu, W., Dandy, G.C., Maier, H.R., 2014. Wu W., Dandy G.C. and Maier H.R. 2014. Protocol
for developing ANN models and its application to the assessment of the quality of the
ANN model development process in drinking water quality modeling. Environ. Model.
Softw. 54:108-127.
Wynne, T.T., Stumpf, R.P., Tomlinson, M.C., Fahnenstiel, G.L., Dyble, J., Schwab, D.J., Joshi,
S.J., 2013. Evolution of a cyanobacterial bloom forecast system in western Lake Erie:
Development and initial evaluation. J. Gt. Lakes Res., Remote Sensing of the Great
Lakes and Other Inland Waters 39(Supplement l):90-99.
Yang, L., Kim, D., Uzun, H., Karanfil, T., Hur, J., 2015. Assessing trihalomethanes (THMs) and
N-nitrosodimethylamine (NDMA) formation potentials in drinking water treatment plants
using fluorescence spectroscopy and parallel factor analysis. Chemosphere 121:84-91.
Yuan, L.L., Pollard, A.I., Pather, S., Oliver, J.L., D'Anglada, L., 2014. Managing microcystin:
identifying national-scale thresholds for total nitrogen and chlorophyll a. Freshw. Biol.
59(9):1970-1981.
Zamyadi, A., Dorner, S., Ndong, M., Ellis, D., Bolduc, A., Bastien, C., Prevost, M., 2014.
Application of in vivo measurements for the management of cyanobacteria breakthrough
into drinking water treatment plants. Environ. Sci. Process. Impacts 16(2):313-323.
Zamyadi, A., McQuaid, N., Prevost, M., Dorner, S., 2012. Monitoring of potentially toxic
cyanobacteria using an online multi-probe in drinking water sources. J. Environ. Monit.
14(2):579-588.
67

-------
Zhang, G.S., Chen, L.L., Chen, J., Ren, Z.M., Wang, Z.J., Chon, T.-S., 2012. Evidence for the
stepwise behavioral response model (SBRM): The effects of carbamate pesticides on
medaka (Oryzias latipes) in an online monitoring system. Chemosphere 87:734-741.
Zhang, G.S., Chen, L.L., Liu, Y.D., Chon, T., Ren, Z.M., Wang, Z.J., Zhao, J.P., Zhao, Y.Y.,
2011. A new online monitoring and management system for accidental pollution events
developed for the regional water basin in Ningbo, China. Water Sci. Technol.
64(9): 1828-1834.
Zhao, H., Hou, D., Huang, P., Zhang, G., 2015. Periodic pattern extraction and anomaly
detection for free chlorine in drinking water network. Water Sci. Technol. Water Supply
15(3): 541—551.
Zhou, Y., Jeppesen, E., Zhang, Y., Shi, K., Liu, X., Zhu, G., 2016. Dissolved organic matter
fluorescence at wavelength 275/342 nm as a key indicator for detection of point-source
contamination in a large Chinese drinking water lake. Chemosphere 144:503-509.
68

-------
Appendix A - Detailed Case Studies
Appendix A Table of Contents
A 1. Danube and Tisza River Basin Case Study	70
A 2. Delaware Valley Case Study	75
A 3. Great Lakes Case Study	80
A 4. Lake Huron to Lake Erie Corridor Case Study	84
A 5. Lower Mississippi River Basin Case Study	88
A 6. Nile River Basin Case Study	92
A 7. Ohio River Basin Case Study	95
A 8. Susquehanna River Basin Case Study	99
A 9. Upper Mississippi River Basin Case Study	103
69

-------
A 1. Danube and Tisza River Basin Case Study
Location
Danube and Tisza River Basin
Source Water
Danube and Tisza River Watershed
Coordinating
International Commission for the Protection of the Danube River
Agency
(ICPDR)
Start Date
1997
System Description
The Danube River runs 1,775 miles (2,857 km) from Germany to the Black Sea. The Danube
River Basin includes the watersheds of multiple tributaries, the largest of which is the Tisza
River (Appendix Figure 1). Home to over 81 million people, the region spans 309,447 square
miles (801,463 square kilometers) in Albania, Austria, Bosnia-Herzegovina, Bulgaria, Croatia,
Czech Republic, Germany, Hungary, Italy, Macedonia, Moldova, Montenegro, Poland, Romania,
Serbia, Slovakia, Slovenia, Switzerland and Ukraine (ICPDR, 2011).
• Prag
50 100 150 200 250 km
Slovak
Bratislava
Munchen
I old ova)
\ Chisinaii1
Tisza River Basin
Switzerlani
Slovenia J V
.	I Zagreb
Ljubljana^ฎ
Sofia
acedonia
Albania
Appendix Figure 1. Danube and Tisza River Basins (WWF, 2002).
The Tisza River runs 600 miles (966 km) from Ukraine to Serbia, where it meets the Danube
River. The Tisza River Basin is home to approximately 14 million people and spans 60,690
square miles (157,186 square kilometers) in Romania, Hungary, Slovakia, Ukraine and Serbia
70

-------
(VRIC & EI, 2014). The Danube River and its tributaries are valuable resources for drinking
water, transportation, agriculture, industry and recreation.
The Danube Accident Emergency Warning System (DAEWS) was implemented in 1997 in
Austria, Bulgaria, Czech Republic, Croatia, Germany, Hungary, Romania, Slovakia and
Slovenia. In 1999, the system was extended to Ukraine and Moldova and in 2005 the expansion
grew to include Bosnia-Herzegovina and Serbia (ICPDR, 2016). System operation, maintenance
and upgrades have been managed by the International Commission for the Protection of the
Danube River (ICPDR) Secretariat (ICPDR, 2015a). In each of the member countries, a Principal
International Alert Center (PIAC) ("Alert Center") was established to manage communications.
The DAEWS is operated as a partnership and communications network between the Alert
Centers, with overall management from the ICPDR. The Alert Center in each country relies on
the national AEWS, which is separate from the DAEWS and managed solely on a national level
(ICPDR, 2014).
"Appropriate control of accidental
pollution is essential in order to
mitigate adverse effects of hazardous
substances spills... The Accident
Emergency Warning System (AEWS)
was developed to... recognize
emergency situations. It is activated if a
risk of transboundary water pollution
exists and alerts downstream countries
with warning messages in order to help
national authorities to put safety
measures ...into action. " (ICPDR,
2015a).
quality monitoring network, (3) an international communication and alert system and (4) a web
and database portal. The proposed Tisza River EWS would include continuous water quality
monitoring with real-time data transmission (VRIC & EI, 2014).
Monitoring Network
The water quality monitoring component of the DAEWS is referred to as the TransNational
Monitoring Network (TNMN), a network of water quality monitoring programs in member
countries. Illustrated in the Appendix Figure 2, the TNMN includes 79 monitoring stations and
measures 52 water quality parameters at least 12 times per year (ICPDR, 2016; VRIC & EI,
2014). In addition to the periodic monitoring of the TNMN, the DAEWS relies on water quality
monitoring and incident notification through the national Accident Emergency Warning System
(AEWS) of each country with Alert Centers as the central points of the communication network.
In addition to the DAEWS, an early warning
system (EWS) is being considered for the Tisza
River Basin. A study was conducted by the
Veszpremi Regionalis Innovacios Centrum
Nonprofit Kft. and Environmental Institute to
(VRIC & EI, 2014):
•	Consider monitoring locations and
constituents of interest
•	Assess monitoring equipment options
•	Examine existing EWS applications
•	Develop anticipated capital and
operational costs
The existing DAEWS consists of: (1) a
partnership of stakeholders, (2) a periodic water
71

-------
Transnational Monitoring Network - Surface Waters
DRBM Plan - Update 2015 - MAP 19
Praha
Stuttgart

Dona**
Miskolc
Chifinau
Budapest1'
Kecskemet
Ratra
Kanal Dunav-l - Timisoa ra
Tha-Dunav \ ' e\ jO-
Lacuf ~
Razm*\
LEGEND
Surface Water Body Monitoring Stations:
Beograd
SERBIA
J 10OO:0OOinhafcrtarts
~	Danube River Basin District
™	Danube River
—	Tributaries ('with catchment area > 4,000 km^
~	Lake water bodies (with surface area >100
ฆ	Transitional water bodies
ฆ	Coastal water bodies
—	Canals
—	National borders
Podgori<
•StrveHsrwe UonXofttg 1 pttMoes an assess-new of trie oc&af surface rater ssrus m tte Danoce River Basv OSKt
"Swveflance Utntto/ttg 2 provides an assessTOftf cr cop-fen? f/axSs otspecticpoNutanis and cf eafls of sifistsrces rrareterretf dew^-Srsam We c&ntA&s
ป. tcpdr.org
icpdr
Vienna. December 2015
Appendix Figure 2. Danube River transnational monitoring network (ICPDR, 2015b).
72

-------
Data Management and Communication
The DAEWS includes a website with tracking of incident reports and alerts and a database of
chemical information.
Communication and Response Network
The basis of the communication and response network are the Alert Centers. The DAEWS is
triggered when an Alert Center is notified of a pollution incident through the national AEWS.
The primary tasks of the Alert Centers are communication, assessment and decision-making.
Incidents are reported via the DAEWS website with email and text notifications to affected
parties, including downstream Alert Centers. The notification network is operational 24-
hour s/day, with telephone backup.
Future Development
As of the 2014 report, the proposed addition of an EWS for the Tisza River would proceed as
follows (VRIC & EI, 2014):
•	Feasibility study in 2014
•	Pilot proj ect in 2014 through 2015
•	EWS basic basin-wide development in 2016 through 2018
•	EWS expansion and additions in 2019
Challenges, Insights and Operational Notes
The DAEWS enables communication across national boundaries between many countries of
varying governing structures, priorities, cultures and economies. Essential aspects include
agreement between countries within the Danube River Basin and standardization of operations,
reporting and communication. The Alert Centers are the core component of the system,
functioning as the connection between national authorities and international communications.
Contact Information
ICPDR Secretariat, Vienna International Centre, Wagramer Strasse 5, A-1220 Vienna, Austria,
+431 260 60 5738, icpdr@unvienna.org.
References
ICPDR, International Commission for the Protection of the Danube River. (2011). Danube Basin
Facts and Figures. Retrieved February 1, 2016, from https://www.icpdr.org/main/danube-
basin/countries-danube-river-basin.
ICPDR, International Commission for the Protection of the Danube River. (2014). International
Operations Manual for PIACs of the Danube AEWS. Retrieved February 1, 2016, from
https://www.icpdr.org/main/activities-projects/aews-accident-emergency-warning-system
ICPDR, International Commission for the Protection of the Danube River. (2015a). The Danube
River Basin District Management Plan, Part A - Basin-wide overview. Retrieved January
21, 2016, from https://www.icpdr.org/main/management-plans-danube-river-basin-
published
ICPDR, International Commission for the Protection of the Danube River. (2015b).
Transnational Monitoring Network - Surface Waters. Retrieved February 1, 2016, from
73

-------
http://www.icpdr.org/main/publications/maps-danube-river-basin-district-management-
plan-2015
ICPDR, International Commission for the Protection of the Danube River. (2016). AEWS -
Accident Emergency Warning System. Retrieved February 1, 2016, from
https://www.icpdr.org/main/activities-projects/aews-accident-emergency-warning-system
VRIC & EI, Veszpremi Regionalis Innovacios Centrum Nonprofit Kft., and Environmental
Institute. (2014). Water Quality Early Warning System - On Transboundary
Watercourses of the Tisza River Basin. Retrieved December 24, 2015, from
http://www.danubewaterquality.eu/uploads/mod_files/WQM-EWS_part-l-
4_EN_v2.0_.pdf
WWF, World Wildlife Fund. (2002). The Ecological Effects of Mining Spills in the Tisza River
System in 2000. Retrieved February 1, 2016, from
https://assets.panda.org/downloads/Tisza_Cyanide_Report.pdf
74

-------
A 2. Delaware Valley Case Study
Location
Delaware Valley
Source Water
Delaware and Schuylkill River Watersheds
Coordinating
Philadelphia Water Dept., EWS Steering Committee, Advisory
Agency
Committee
Start Date
2004
System Description
The Delaware River runs 330 miles from Hancock, New York to the Delaware Bay. The
Delaware River Basin is comprised of the Schuylkill River and Delaware River watersheds
(Appendix Figure 3). Home to approximately 8 million residents, the region spans 13,500 square
miles in parts of Pennsylvania, New Jersey, New York and Delaware (DRBC, 2013). The two
rivers and their tributaries are not only valuable resources as drinking water supplies, but also for
transportation, agriculture, industry and recreation.
"The EWS provides advanced
warning of water quality events,
web-based tools for determining
proper event response, and a
strong partnership between water
users and emergency responders in
the Schuylkill and Delaware River
watersheds " (Philadelphia Water
Department, 2008).
Philadelphia Source Watersheds
Delaware River Watershed
Schuylkill River Watershed
Pe
New
New York City
Pa
Allentown
Trenton
Philadelphia
Appendix Figure 3. Delaware Valley water basin
(Anderson, 2015).
75

-------
As part of the region's Source Water Protection Program, the Philadelphia Water Department led
the development of the Delaware Valley Early Warning System to enable advanced notification
to water users of contamination incidents and changes in water quality. Implemented in 2004,
the system was developed and is operated as a partnership between 300 water users in 50
organizations. Members include 25 water treatment plants (21 plants from 13 utilities in
Pennsylvania and four plants from four utilities in New Jersey), 24 industrial sites (in
Pennsylvania, New Jersey and Delaware), the Pennsylvania Department of Environmental
Protection (PA DEP), the New Jersey Department of Environmental Protection (NJ DEP), the
Delaware River Basin Commission (DRBC), U.S. EPA, U.S. Geological Survey (USGS), U.S.
Coast Guard (USCG), county offices of emergency management and county health departments
(Anderson, 2015). The governing body of the system is the EWS steering committee, which is
comprised of utility representatives. Government agency representatives participate through an
advisory committee. The funding structure for the system is based on a user fee from drinking
water utility and industry members. Highlighting the success of the system, the Delaware Valley
EWS received the 2015 Pennsylvania Governor's Award for Environmental Excellence
(Philadelphia Water Department, 2015). The Delaware Valley EWS consists of four key
elements: (1) a partnership of stakeholders, (2) a monitoring network, (3) a notification system
and (4) a web and database portal.
Monitoring Network
The system includes two monitoring pathways: (1) telephone and web-based incident reporting
and (2) water quality monitoring stations throughout the watershed. Contamination incidents are
reported by emergency personnel, water system representatives, or other EWS members through
a web-based form or via telephone. The incident time, location, and additional details are
recorded, which triggers automatic notification procedures. Water quality data are collected
every 15 minutes from 88 locations (Appendix Figure 4) including drinking water treatment
plant intake locations and USGS monitoring stations, with remote data transmission to the
system's server (Duzinski, 2008). At drinking water treatment plant intake locations, real-time
monitoring equipment measures dissolved oxygen (DO), turbidity, temperature and conductivity
(Duzinski, 2008). The current EWS coverage area includes 121 intakes, 280 miles of river, 7,400
miles of stream and 88 water quality monitoring stations (Anderson, 2015).
76

-------
IULLIVAN
ORANGE COUNTY
PIKE COUNTY
:OUNTY
STER COUNTY
CฉJNTY

Industrial Intake
<ง>
Power Intake
<ง>
Water Supply Intake
•
Monitoring Station
~
EWS Coverage Area
•	121 Intakes
•	88 Monitoring Stations
•	Web-Based Reporting
•	Telephone Reporting
Partnership between:
•	25 Water Treatment Plants
•	24 Industrial Sites
•	300 Water Users from
•	50 Organizations
Adapted from Anderson (2015)
Appendix Figure 4. Delaware Valley monitoring stations (adapted from Anderson, 2015).
Data Management and Interpretation
The Delaware Valley EWS database manages both reported incident data and water quality
monitoring data. The database is accessible through a secure web portal. Water quality data are
available for every 15 minutes of the preceding 30-day period and as a daily average dating back
several years. Additional water quality information, collected through standard water treatment
plant operations, is logged as well. The portal includes online tools to access water quality and
incident data and to view and analyze water quality data. Historical USGS flow data are
incorporated into the time-of-travel model to predict arrival times at downstream locations
(Gullick et al., 2004). In addition to incident assessment and water quality data analysis, the web
portal enables spill model analysis and integrated mapping. Most recently, the EWS has
developed a tidal model to assess the impact of tidal flow in the Lower Delaware River.
Communication and Response Network
The system is fully automated such that incident reporting triggers the time of travel model and
automatically notifies system users, without the need for 24-hour staffing. Timely notification of
incidents is facilitated by consistent formatting of incident reporting. The automated incident
77

-------
reporting and notification system has the ability to quickly provide warning to a large area,
allowing for expedited response.
Future Development
The Delaware Valley EWS was designed as a framework capable of expansion and perpetual
enhancement. Since the system became operational in 2004, numerous improvements have been
made and upgrades and advancements are ongoing. Anderson (2015) identified the future plans
as:
•	Monitoring for additional parameters
•	Increasing monitoring locations
•	Technology upgrades
•	Expansion of user base and geographical coverage
•	Enhancement of web-based geospatial and modeling tools
Challenges, Insights and Operational Notes
An essential element of the system's success is the partnership approach encouraging active user
participation. With member participation in the EWS steering committee, member priorities can
be addressed with the appropriate guidance of the advisory committee.
The automated incident reporting and notification system allows for prompt and wide-reaching
dissemination of time-sensitive information, without the need for 24-hour staffing.
Regional water consumer participation and interest in the Delaware Valley EWS has been
promoted by the potential adaptability of the developed network for other applications (e.g.,
Department of Health interest in application of the network as a recreational waterborne disease
tracking tool) (Gullick et al., 2004).
Contact Information
Philadelphia Water Department, Source Water Protection Program Manager, Kelly Anderson,
(215) 685-6245, kelly.anderson@phila.gov.
References
Anderson, K. (2015). Delaware Valley Early Warning System. 2015 Exchange Network
National Meeting, Philadelphia, PA. September 29-October 1, 2015. Retrieved December
24, 2015, from http://www.exchangenetwork.net/EN2015_files/9.3.pdf
DRBC, Delaware River Basin Commission. (2013). Basin Information. Retrieved January 18,
2016, from http://www.nj.gov/drbc/basin/
Duzinski, P. (2008). Delaware Valley Early Warning System (EWS) - Regional Response Team
III, Pittsburgh, PA. September 24, 2008. Retrieved December 24, 2015, from
http://www.nrt.org/production/NRT/RRT3.nsf/Resources/powerpointSep08-
2/$File/PDuzinski.pdf
Gullick, R. W., Gaffney, L. J., Crockett, C. S., Schulte, J., & Gavin, A. J. (2004). Developing
regional early warning systems for U.S. source waters. J. Am. Water Works Assoc. 96(6), 68.
Philadelphia Water Department. (2008). Delaware Valley Early Warning System - EWS fact
sheet. Retrieved December 24, 2015, from
http://www.schuylkillwaters.org/doc_files/EWSfactsheet.pdf
78

-------
Philadelphia Water Department. (2015). Philadelphia Water's Early Warning System Getting
Praise from High Places. Retrieved January 2, 2016, from
http://www.phillywatersheds.org/philadelphia-waters-early-warning-system-getting-praise-
high-places
79

-------
A 3. Great Lakes Case Study
Location
Lake Erie
Source Water
Coordinating
Agency
Start Date
Lake Erie
NOAA Great Lakes Environmental Research Laboratory (GLERL)
1989 (Starting date for the Cooperative Institute for Limnology and
Ecosystems Research)
System Description
The Great Lakes Harmful Algal Blooms (HABs) Program is a collaborative effort between
scientists at National Oceanic and Atmospheric Administration (NOAA), Great Lakes
Environmental Research Laboratory (GLERL) and the Cooperative Institute for Limnology and
Ecosystems Research (CILER). Project goals are to provide a five-day prediction of the severity
and movement of HABs on Lake Erie. Elements of the effort are water quality monitoring,
including near real-time microcystin concentration monitoring, collection and analysis of
satellite and hyperspectral data, modeling and forecasting and communication of results to the
general public via web tools and a periodic Lake Erie HAB bulletin.
Monitoring Network
Satellite data, in conjunction with remote sensing buoys, and a comprehensive physical
monitoring program are used to forecast HABs. From 2002 through 201 1, medium-spectral
resolution imaging spectrometer's (MERIS's) satellite data was use to model cyanobacterial
blooms (Stumpf et al., 2012). Starting in 2012, MERIS lost communication and could no longer
be used, so Moderate Resolution Imaging Spectrometer's (MODIS's) satellite data has been used
since that time. The physical sampling locations are mapped in Appendix Figure 5. Four remote
sensing buoys collect the following data every 15 minutes (NOAA GLERL, 2016):
•	Air temperature
•	Water temperature
•	Barometric pressure
•	Wind speed
•	Wind direction
•	Turbidity
•	Conductivity
•	Phosphorus
•	Chlorophyll-a
•	Blue-green phycocyanin
•	Fluorescent decomposed organic matter
The following data are collected at the surface, 0.75 meters below the surface and 1 meter above
the bottom of the lake:
80

-------
•	Particulate microcystin
•	Dissolved microcystin
•	Chlorophyll-a
•	Phycocyanin
•	Temperature
Western Lake Erie
Master Stations
Lake Erie
WE 14
WE 15
10 Kilometers
Appendix Figure 5. Western Lake Erie sample locations.
Data Management and Interpretation
Unprocessed satellite data is gathered from the NASA MODIS - Terra and aqua satellites, and is
then modeled into cyanobacteria biovolume using the second derivative spectral shape algorithm
(Wynne et aL 2008). This algorithim uses a change in the shape of the spectral curve at 681 nm
to distinguish cyanobacterial blooms from algal growth. The fly over time for the MODIS
satellites is every one to two days. Real-time monitoring data is collected from the solar-powered
bouys and transmitted by a cellular network. Weekly physical samples are collected by
University of Michigan in conjuction with NOAA GLERL.
Communication and Response Network
A forecast bulletin is issued up to every two days during the bloom season to subscribers and is
available online to the general public. The online FIAB tracker is updated daily with a five-day
81

-------
forecast. All of the real-time, field measurements, laboratory data and bulletins are publically
available online, along with the satellite images.
Future Development
Satellite images are limited to a water depth of 1 meter. Wind can push cyanobacterial cells
lower in the water column, which causes inaccuracy in forecasting (Wynne et al., 2010). A three-
dimensional model is being developed that will account for the impact of the wind. The model
will be in experimental trials during 2016 (personal communication Timothy Davis, NOAA
GLERL, January 27, 2016). This model will help drinking water managers to determine at what
depth to expect cyanobacteria and will also help managers to improve the forecasting capabilities
of HAB tracker.
An experimental model will be developed based on data from the environmental sample
processer on the buoys, with a goal of forecasting the toxicity of the bloom.
The Cyanobacteria Assessment Network (CyAN) is a collaboration between NOAA, USGS and
U.S. EPA (U.S. EPA, 2016). Remote sensing data will be produced by the Sentinel-2, Sentinel-3,
and Landsat satellites (Schaeffer et al., 2015). Data collected from Sentinel-3 is of particular
interest because it is equipped with a sensor that is expected to have better sensitivity and
resolution than the MODIS or MERIS sensors (Lunetta et al., 2015). The sensor on the Sentinel -
3 is called the Ocean Land Colour Instrument (OLCI). Sentinel-3 was launched in February of
2016, and has begun producing data. Data from Sentinel-3, processed through the model that has
been developed for remote sensing in Lake Erie, will be used in Ohio, California, Florida, New
Hampshire, Massachusetts, Connecticut and Rhode Island.
Challenges, Insights and Operational Notes
The estimated threshold for cyanobacteria detection is 20,000 cells/mL (NOAA Great Lakes
Environmental Research Laboratory, 2015b). This is the same value as the WHO guideline level
for recreational waters, which is based on the skin irritation impacts, and not the toxicity of
microcystin that is ingested (WHO, 2003). A count of 20,000 cells/mL is in the high alert level
range for drinking water sources (Newcombe et al., 2010). The alert levels are summarized in
Appendix Table 1. Drinking water alert levels are lower than the recreational alert levels because
they account for the toxicity of ingested microcystin. For drinking water forecast and early
warning, it would be ideal to have a lower detection threshold.
Appendix Table 1. Drinking Water Cyanobacteria Alert Levels (adapted from Newcombe
et al., 2010)
Alert level
Cell count
Low alert
500 to 1,999 cells/mL of cyanobacteria
Medium alert
2,000 to 6,499 cells/mL of Microcystis aeruginosa
High alert
6,500 to 64,999 cells/mL of Microcystis aeruginosa
Very high alert
More than 65,000 cells/mL of Microcystis aeruginosa
Newcombe et al., 2010. Management Strategies for Cyanobacteria (blue-green algae): A Guide for Water Utilities.
Water Quality Research Australia Limited: Adelaide, Australia.
A study using MERIS was able to demonstrate detection as low as 10,000 cells/mL (Lunetta et
al., 2015). This detection level, although an improvement, is still in the high alert range.
82

-------
Contact information
Website: http://www.glerl.noaa.gov/res/HABs_and_Hypoxia/
Timothy Davis, Ph.D., Molecular HAB Ecologist, NOAA, 734-741-2286,
Timothy.Davis@noaa.gov
Rick Stumpf, Ph.D., Oceanographer, NOAA, 301-713-3028 \173, Richard.Stumpf@noaa.gov
References
Lunetta, R.S., Schaeffer, B.A., Stumpf, R.P., Keith, D., Jacobs, S.A., Murphy, M.S., 2015.
Evaluation of cyanobacteria cell count detection derived from MERIS imagery across the
eastern U.S.A. Remote Sens. Environ. 157:24-34.
Newcombe, G., House, J., Ho, L., Baker, P., Burch, M., 2010. Management Strategies for
Cyanobacteria (blue-green algae): A Guide for Water Utilities, (Cooperative Research
Centre for Water Quality and Treatment [Australia]; Research Report No. 74). Water
Quality Research Australia Limited: Adelaide, South Australia.
NOAA GLERL, 2016. NOAA Great Lakes Environmental Research Laboratory (GLERL) Great
Lakes Water Quality [WWW Document], URL
http://www.glerl.noaa.gov/res/HABs_and_Hypoxia/ habsTracker.html (accessed January
26, 2016)
NOAA GLERL, 2015. Experimental Lake Erie Harmful Algal Bloom, Bulletin 17. Ann Arbor,
MI: NOAA Great Lakes Environmental Research Laboratory.
Schaeffer, B., Loftin, K., Stumpf, R., Werdell, P., 2015. Agencies collaborate, develop a
cyanobacteria assessment network. Eos 96, doi: 10.1029/2015EO038809.
Stumpf, R.P., Wynne, T.T., Baker, D.B., Fahnenstiel, G.L., 2012. Interannual variability of
cyanobacterial blooms in Lake Erie. PLoS ONE 7:e42444.
U.S. EPA, 2016. Cyanobacteria Assessment Network (CyAN) Project [WWW Document], URL
http://www.epa.gov/water-research/cyanobacteria-assessment-network-cyan-project
(accessed January 28, 2016).
WHO, 2003. Guidelines for safe recreational water environments Volume 1 : Coastal and fresh
waters. WHO: Geneva, Switzerland.
http://www.who.int/water_sanitation_health/bathing/srwe 1 execsum /en/
index7.html
Wynne, T.T., Stumpf, R.P., Tomlinson, M.C., Warner, R.A., Tester, P.A., Dyble, J., Fahnenstiel,
G.L., 2008. Relating spectral shape to cyanobacterial blooms in the Laurentian Great
Lakes. Int. J. Remote Sens. 29:3665-3672.
Wynne, T.T., Stumpf, R.P., Tomlinson, M.C., Dyble, J., 2010. Characterizing a cyanobacterial
bloom in Western Lake Erie using satellite imagery and meteorological data. Limnol.
Oceanogr. 55, 2025-2036.
83

-------
A 4. Lake Huron to Lake Erie Corridor Case Study
Location
Lake Huron to Lake Erie Corridor, Michigan
Source Water
U.S. side of St. Clair River, St. Clair Lake, Detroit River
Coordinating
Michigan Department of Environmental Quality
Agency

Start Date
2006
System Description
The St. Clair River connects Lake Huron with Lake St. Clair, which is connected to Lake Erie
via the Detroit River (Appendix Figure 6). This high traffic corridor is important as a shipping
route and a popular recreational area. The west side of the Huron to Erie corridor is in the U.S.
(Michigan) while the east side is in Canada (Ontario). The Huron to Erie corridor serves as a
water supply for over four million people (Howard, 2007).
Lake
Huron
CANADA
St. Clair
River
~
l"d. Ohio
CANADA
Lake St
Ann Detroit C/a/r _
Arbor	~ — - lima no
Windsor
Detroit 'ih -Fighting Island
River^ r
Lake Erie
"It is our opinion that the Huron
to Erie Network is one of the best
tools available to maintain safe
drinking water " (City of
Marysville, Wrubel, 2014).
Appendix Figure 6. Lake Huron to Lake Erie corridor (Morrison, 2006).
The Huron-to-Erie Real-time Drinking Water Protection Network is a spill detection and
notification system for water suppliers along the corridor, with near-real-time monitoring data
and instantaneous notification. Starting in 2006, this EWS was developed through a partnership
84

-------
between U.S. EPA, the U.S. Department of Homeland Security, the Michigan Department of
Environmental Quality, the health departments of Macomb and St. Clair Counties, and 14
drinking water treatment plants (Wrubel, 2014; Lichota and DeMaria, 2009). Funding for the 14
sites was initially received from government grants. The current coverage area includes nine
monitoring sites. The decrease in monitoring sites is attributed to limited funding, which is
currently supplied through water rates (Wrubel, 2014).
The Huron-to-Erie Real-time Drinking Water Protection Network consists of (1) a water quality
monitoring network, (2) a web and database portal and (3) a notification system.
Monitoring Network
Water quality data are collected from nine monitoring stations, located at drinking water
treatment plants along the Huron to Erie corridor as illustrated in Appendix Figure 7; plant
names in red are no longer included in the network (Wrubel, 2014). The network is equipped
with NexSens 5100-iSIC Data Loggers (Fondriest Environmental, Fairborn, OH), 6600 V2-4
Multi-Parameter Water Quality Sondes sensor (YSI Inc., Yellow Springs, OH), Turner
Flydrocarbon Fluorometers, I lapsite' ER gas chromatograph - mass spectrometers (GC-MS)
(Inficon, Bad Ragaz, Switzerland) and Hachiv (Hach, Loveland, CO) total organic carbon (TOC)
analyzers (NexSens, 2016).
With near-real-time data transmission, water quality data are logged every 15 to 30 minutes. The
monitoring equipment measures pH, temperature, DO, conductivity, turbidity, oxidation
reduction potential, chlorophyll, organic carbon, gasoline, diesel fuel, waste oils and other
industrial chemicals.
(m
'v>„
Marine City
Algonac
Mt. Clemens 1
Detroit-WW Park
Southwest Detroit
Wyandotte
I Park'**'
t
Monroe i
Appendix Figure 7. Lake Huron to Lake Erie drinking water monitoring network (Wrubel,
2014).
85

-------
Data Management and Interpretation
The Huron-to-Erie Real-time Drinking Water Protection Network database manages water
quality monitoring data that is accessible through a secure web portal. Using NexSens iChart
software, the online data portal includes tools to view and analyze water quality monitoring data.
Interpretation of data includes delineation of alarm conditions, which will vary based on
parameter. For example, if industrial chemical detection exceeds either 10%, 50% or 90% of the
allowable level, it would be categorized as either an anomaly, a potential spill or a likely spill,
respectively (Lichota and DeMaria, 2009). A secondary website for public access to water
quality data was also developed.
Communication and Response Network
The system software includes a communication and data sharing network. Based on alarm
configuration, as discussed above, water suppliers are notified of adverse water quality
conditions for prompt adjustment to plant operations.
Future Development
Potential future plans for the Huron to Erie Drinking Water Protection Network include (Wrubel,
2014):
•	Coordination with Canadian agencies to enhance warning system
•	Improvement of the flow model
•	Integration of incident reporting
•	Public outreach and education
Challenges, Insights and Operational Notes
In the configuration of alarm levels, it is important to integrate background levels for proper spill
detection (Lichota and DeMaria, 2009). The system has been challenged by the availability of
sufficient, sustainable funding and regional organizational coordination. The loss of multiple
monitoring stations in the network can be detrimental to the robustness of the EWS.
Contact Information
Michigan Department of Environmental Quality, Brock F. Howard, (517) 335-4101,
howardb 1 @mi chigan. gov.
City of Marysville, Supervisor of Water/Wastewater Operations, Bari Wrubel, 1535 River Rd.
Marysville, MI 48040, (810) 364-8460, bwrubel@cityofmarysvillemi.com.
References
Howard, B. (2007). Real Time Monitoring Program - Protecting the Drinking Water Source in
the Huron to Erie Corridor. Retrieved January 2, 2016, from
http://www.michigan.gov/documents/deq/deq-wb-wws-BrockHowardCIPRTMtalk5-30-
07_237078_7.pdf
Lichota, S. and DeMaria, A. (2009). St. Clair River - Lake St. Clair - Detroit River, Drinking
Water Protection Network. Retrieved January 21, 2016, from
http://www.allianceofrougecommuni ties. com/PDF s/SWAG/DWPP%20May%202009%20S
WAG%20Presentation.pdf
86

-------
Morrison, S. (2006). Lake Whitefish Returning to the Detroit River to Spawn; Federal Scientists
Document First Reproducing Population of Whitefish in the River Since 1916. United States
Geological Survey. Retrieved January 22, 2016 from,
http://soundwaves.usgs.gov/2006/08/research.html
NexSens Technology. (2016). St. Clair & Detroit River Monitoring, Michigan Department of
Environmental Quality. Retrieved January 21, 2016, from
http://www.nexsens.com/case_studies/st-clair-detroit-river-monitoring.htm
Wrubel, B. (2014). Underwater Detective: The Importance of Near-Real Times River Water
Monitoring. Retrieved January 21, 2016, from
http://www.friendsofstclair.ca/www/pdf/2014%20Symposium/Drinking%20Water%
20Intake%20Monitoring%20-%20W rubel. pdf
87

-------
A 5. Lower Mississippi River Basin Case Study
Location
Lower Mississippi River Basin
Source Water
Lower Mississippi River Watersheds
Coordinating
Louisiana Department of Environmental Quality (DEQ)
Agency

Start Date
1986
System Description
The Lower Mississippi River runs 954 miles from Cairo, Illinois to the Gulf of Mexico
(LMRCC, 2014) (Appendix Figure 8). The Lower Mississippi River Basin spans approximately
105,000 square miles and includes portions of Missouri, Kentucky, Tennessee, Arkansas,
Mississippi and Louisiana (USDA, 2013). The Lower Mississippi River is a valuable resource
for drinking water, energy production, shipping, industry and recreation.
"To combat any threat to drinking
water drawn from the river, DEO,
potable water works and industries
along the river entered into a
cooperative agreement in 1986 to
found the Early Warning Organic
Compound Detection System
(EWOCDS) " (Louisiana DEO,
2014).
Appendix Figure 8. Lower Mississippi River Basin (Missouri DNR, 2016).
MISSOURI

LOUISIANA
Mwfl" ft
.MEXICO
LOWER MISSISSIPPI
RIVER BASIN
ILLINOIS
ARKANSAS
88

-------
The Early Warning Organic Compound Detection System (EWOCDS) was implemented in 1986
for the southern-most portion of the Lower Mississippi River, covering Louisiana from Baton
Rouge to Plaquemines Parish (Wold, 2015). Over 350 facilities are located along the Lower
Mississippi River in Louisiana, where over 1.6 million residents rely on the river as a drinking
water supply (Louisiana DEQ, 2014). The original system consisted of nine monitoring stations
and was developed with grant funding from U.S. EPA, with subsequent costs covered by the
Louisiana Department of Environmental Quality (DEQ) (Louisiana DEQ, 2009). By 2009, after
problems with funding and staffing and the withdrawal of some participants from the program,
only six monitoring stations were operating (Louisiana DEQ, 2009). Despite a limited budget,
the system has remained operational, with one DEQ environmental scientist responsible for
management and maintenance. In 2014, the EWOCDS benefitted from a settlement between
Exxon Mobil and Louisiana, from which $250,000 was slated for additions and upgrades to the
EWS, including new computers and gas chromatographs (Wold, 2015; Louisiana DEQ, 2014).
Additionally, funding from the Louisiana Chemical Association was provided to improve
consistency throughout the system with the development of standard operating procedures for all
monitoring stations (Wold, 2015).
The Lower Mississippi River EWOCDS consists of (1) a partnership between stakeholders, (2) a
monitoring network and (3) a notification system.
Monitoring Network
The Louisiana DEQ maintains monitoring equipment, while the sample analysis is the
responsibility of the staff at monitoring locations (Louisiana DEQ, 2014). Water quality data are
collected from seven monitoring stations. Each location includes a gas chromatograph, with
samples collected twice per day at most sites. Two stations have continuous sampling, enabling
real-time water quality monitoring. Monitoring stations measure 28 chemical contaminants
including halogenated organic compounds, chlorinated hydrocarbons, and trihalomethanes.
Results from onsite sample analyses are submitted to the Louisiana DEQ. Monitoring sites as of
2012 are illustrated in Appendix Figure 9, with former sites in gray.
Data Management and Interpretation
Results are reported to the Louisiana DEQ and cataloged in a database. If an incident is detected
at a monitoring location, a second sample is used to confirm contaminants. Results are available
to the public upon request.
Communication and Response Network
The Louisiana DEQ representative notifies water users if an incident has occurred. The
Louisiana DEQ also provides a portal through their website for reporting incidents. Publicly
available information does not indicate how EWOCDS data are used during an incident
response.
Future Development
Potential future plans for the Lower Mississippi River's EWOCDS include (Wold, 2015):
•	System expansion
•	Increased participation
•	Public outreach
•	Website improvements to include public access to water quality summary data
89

-------
James
St. Bernard
Chasse
Map Number 201201036
Date: March 16. 2012
Projection UTM. NAD 83 Zone 15
Source 2007 LDOTD Parish. Census Urbanized Areas;
2012 LDEQ EWOCDS stations: LDEQ Surface Water Intakes
LDEQ Disciamei
The Louisiana Department at Envtrpnmert* QuMy (LDEQ) has nM every reasonable eftwi lo ormjrซ
quoley ana accuracy rn prcdixing 9ss map or data so! Neverthefessne user should be aware thai (he
reformation on whch ll a baaed may have coma from any a vanely of wurces *t*ch an of varying
degrees of map accuracy Therefore LOCO cannot guarantee the accuracy of thป map or data tot aid
does not accept any responsibly for the consequences of its use
EWOCDS
A EWOCDS Station
/\ EWOCDS Station Removed
A Surface Water Intake
Appendix Figure 9. EWOCDS monitoring locations as of 2012 (Louisiana DEQ, 2012).
Challenges, Insights and Operational Notes
A significant challenge of the EWOCDS is that the program is for qualitative screening (absence
or presence) only and not subject to rigorous quality control and quality assurance procedures
due to limits on costs and participant expertise. Like other EWSs, the EWOCDS has a limited
number of funding sources which, in turn, limit the reach and scope of the program.
Contact Information
Louisiana Departm ent of Environmental Quality, Office of Environmental Compliance,
Inspection Division, Tom Killeen, (225) 219-3600, deqinspection@LA.GOV,
http://www.deq.louisiana.gov/
References
LMRCC, Lower Mississippi River Conservation Committee. (2014). Summary of Available
Water Quality Assessments of the Lower Mississippi River. Retrieved January 31, 2016,
from http://www.lmrcc.org/wp-
content/uploads/2015/08/LMRCC_WQ_FlNAL_Sept2014.pdf
Louisiana DEQ, Department of Environmental Quality. (2009). Commission on Streamlining
Government Executive Summary. Retrieved January 30, 2016, from
90

-------
http://www.deq.louisiana.gOv/portal/Portals/0/news/DEQ%20Executive%20Summary%20Re
port%20Commission%20on%20Streamlining%20Government%20081709.pdf
Louisiana DEQ, Department of Environmental Quality. (2012). EWOCDS (Early Warning
Organic Compound Detection System); map. Retrieved January 31, 2016, from
http://map.ldeq.org/images/maps/201201036.jpg Early Warning Organic Compound
Detection System Louisiana DEQ, Department of Environmental Quality. (2014). Early
Warning System helps protect the Mississippi River as a drinking source. Discover DEQ,
Louisiana Department of Environmental Quality Newsletter, November 2014, Issue Number:
34. Retrieved, January 30, 2016, from
http://www.deq.louisiana.gOv/portal/portals/0/news/pdf/DEQENewsletterNovember2014.pdf
Missouri DNR, Department of Natural Resources (2016). Subject: Missouri Geological Survey,
Water Resources, Interstate-waters, Mississippi. Retrieved January 30, 2016, from
http://dnr.mo.gov/geology/wrc/interstate-waters/mississippi_river.htm
USD A, United States Department of Agriculture. (2013). Assessment of the Effects of
Conservation Practices on Cultivated Cropland in the Lower Mississippi River Basin.
Retrieved January 31, 2016, from
http://www.nrcs.usda.gov/Internet/FSE_DOCUMENTS/stelprdbll76978.pdf
Wold, A. (2015). Pollution monitoring program gets boost thanks to money from a settlement
last year between ExxonMobil and Louisiana. The Advocate, August 25, 2015. Retrieved
January 30, 2016, from http://theadvocate.com/news/13203890-123/settlement-funds-
to-upgrade-mississippi
91

-------
A 6. Nile River Basin Case Study
Location
Nile River Basin, Egypt
Source Water
Coordinating
Agency
Start Date
Nile River Watershed
Ministry of Water Resources and Irrigation, National Water Research
Center
2008
System Description
The Nile River runs 4,160 miles from the Kagera Basin to the Mediterranean Sea (Nile Basin
Initiative, 2016). Spanning approximately 1.24 million square miles, the Nile River Basin is
home to 238 million people and includes portions of 11 countries including Burundi, Democratic
Republic of the Congo, Egypt, Ethiopia, Eritrea, Kenya, Rwanda, South Sudan, Sudan, Tanzania
and Uganda (Nile Basin Initiative, 2016) (Appendix Figure 10). The Nile River is a valuable
resource for drinking water supply, agriculture, transportation, recreation and industry.
92

-------
25f
Mediterranean Sea
NILE RIVER BASIN
O SELECTED CITIES
NATIONAL CAPITALS
	MAJOR ROADS
JORDAN

LIBYA
ARAB REP.
OF EGYPT
SAUDI
ARABIA
CHAD
ERITREA;
ASMARA*#
KHARTOUM r
REP. OF
YEMEN
SUDAN
A! Ubayyid O
DJIBOUTI
CENTRAL
AFRICAN
REPUBLIC
^Mongolia
SOMALIA
KENYA
RWANDA-
/ lake Ki"t
INDIAN
OCEAN
BUJUMBURAฎ
Shinyonga
Ofabora OSingindo
TANZANIA
"The implemented real time water
monitoring and reporting system
allows senior water managers to
protect the integrity of Egypt's vital
water resources, as well as, report
the suitability of water for
designated beneficial water uses."
(Khan et ah, 2011).
Appendix Figure 10. Nile River Basin (World Bank, 2000).
The Nile River Basin Early Warning System was developed with funding through North Atlantic
Treaty Organization's (NATO's) Science for Peace Program with coordination from Egypt's
Ministry of Water Resources and Irrigation, National Water Research Center. The EWS consists
of (1) a monitoring network and (2) an internal database portal.
Monitoring Network
The monitoring network consists of eight monitoring sites along the Nile River in Egypt.
Monitoring stations include the following equipment: Hach Hydrolab multi-parameter probe,
data logger, weather station and a potentiometer. Real-time water quality monitoring equipment
measures pH, DO, temperature, conductivity, ammonia and nitrate at 15 minute intervals. The
data are collected with an automated data retrieval system hourly (Khan et al., 2011).
93

-------
Data Management and Interpretation
Data retrieved hourly from real-time monitoring stations are collected, stored and analyzed in
order to produce graphs for assessing water quality changes. Data are accessible through an
internal web portal.
Communication and Response Network
Based on the information in Khan et al. (2011), the system includes near real-time data
communication. The users can access the data portal to assess water quality changes and take any
necessary corrective action.
Challenges, Insights and Operational Notes
Prior to the implementation of this EWS, water quality monitoring on the Nile River in Egypt
consisted primarily of grab samples which, while effective, are not conducive to prompt
detection and response (Khan et al. 2011). The implemented EWS has the potential for multiple
applications as a tool for integrated water resources management in the Nile River Basin.
Future
Current water related initiatives of the Nile Basin Initiative appear focused on flood, drought and
watershed management.
Contact Information
Ministry of Water Resources and Irrigation, National Water Research Center, Shaden Abdel-
Gawad, Cairo, Egypt, shaden@nwrc-eg.org.
References
Khan, H. Dawe, P., Paterson, R., and Abdel-Gawad, S. (2011). Water Resources Management
System for the Nile River. 2011 International Conference on Environment Science and
Engineering, Bali Island, Indonesia, April 1-3, 2011. Retrieved January 22, 2016, from
http://www.ipcbee.com/vol8/70-S10025.pdf
Nile Basin Initiative. (2016). Understanding the Nile Basin. Retrieved January 22, 2016, from
http://www.nilebasin.org/index.php/about-us/the-river-nile
94

-------
A 7. Ohio River Basin Case Study
Location
Ohio River Basin
Source Water
Ohio River
Coordinating
Ohio River Valley Water Sanitation Commission (ORSANCO)
Agency

Start Date
1977
System Description
The Ohio River runs 98 miles through 20 locks and dams from the confluence of the Allegheny
and Monongahela Rivers in Pittsburgh, Pennsylvania, to the Mississippi River in Cairo, Illinois.
The Ohio River Basin is home to approximately 25 million residents and spans 203,940 square
miles through parts of 11 states (Brosnan, 1999; ORSANCO, n.d.). The river and its tributaries
are not only valuable resources as drinking water supplies, but also for transportation,
agriculture, industry and recreation.
The Ohio River Valley Water Sanitation Commission (ORSANCO) is an interstate commission
with representatives from eight states and the federal government that was created to abate and
prevent pollution in the Ohio River. An overarching goal for ORSANCO is to contribute to
improved water quality for drinking water, industrial water, recreational uses and for maintaining
a healthy and diverse aquatic ecology (Brosnan, 1999). ORSANCO was established in 1948 and
maintains an organics detection system (ODS) to provide its stakeholders with early warning of
water quality incidents on the Ohio River. The ODS began with seven monitoring stations
established in 1977 in response to a large carbon tetrachloride spill on the river. Currently, it
consists of 16 water quality stations at water treatment plants and industries along the river
(Schulte, 2014). Samples from these stations are tested at least daily for 30 analytes (Schulte,
2014). In addition to the ODS, ORSANCO's EWS includes self-reporting from industries and
rivers users, and a communications network of drinking water utilities and industries along the
river (Gullick et al., 2003).
Monitoring Network
The 16 water quality monitoring stations use purge and trap gas chromatographs that are able to
detect and track spills (Gullick et al., 2003) (Appendix Figure 11). The stations use a variety of
detector technologies, including gas chromatography with flame ionization and photoionization-
Hall electrolytic conductivity detectors, to test for the following 30 purgeable organic
compounds (Gullick et al., 2003; Schulte, 2014):
Methylene chloride
1,1 Dichloroethylene
1,1 Dichloroethane
Chloroform
1,1,1 Trichloroethane
Carbon tetrachloride
Benzene
T ri chl oroethyl ene
1,2 Dichloropropane
Dichlorobromomethane
Toluene
T etrachl oroethyl ene
Dibromochloromethane
95

-------
•	Ethylbenzene
•	Chlorobenzene
•	Styrene
•	Bromoform
•	1,3 Dichlorobenzene
•	1,4 Di chlorobenzene
•	1,2 Dichlorobenzene
•	Acrylonitrile
•	1,2 Dichloroethane
•	Trans-1,2 Dichloroethylene
•	Cis-1,3 Dichloropropene
•	Trans-1,3 Dichloropropene
•	Hexachlro-1,3 -butadiene
•	1,1 2,2 tetrachloroethane
•	Trichlorofluoromethane
•	Napthalene
A system upgrade that began in 2009 provided updates to equipment, communications and
software (ORSANCO - Organics Detection System - ODS, n.d.).
ORSANCO
Organics Detection System Installations
NY-
PA
Allegheny R.
Pittsburgl
OH
( Monongahela
Muskingum R.
Wheeling
Great Miami R.
Wabash R.
Scioto R.
Cincinnati
1 Parkersburg
Kanawha R.
Portsmouth
Drinking Water Intakes
Licking R.
Kentucky R.
( Huntington www
:inrU/ R	V V V
insville.
ODS DW/IW intakes
GC/MS capability
Louisville
DW ODS Intakes
Online Process GC
KY
Green R.
VA
ODS sites w/ existing
instrumentation
Cumberland
Paducah'
Tennessee R.
^		/ / /	PA
/	Allegheny R. ^
-	V Pittsburg]
in I 0H	••
ff) i	l Monongahela F
Appendix Figure 11. ORSANCO ODS monitoring locations (Schulte, 2014).
Data Management and Interpretation
ORSANCO's ODS stations follow a common sampling procedures manual and results are
reported on at least a weekly basis (Gullick et aL 2003). The system upgrade that began in 2009
included automated detection notification, updated data management architecture, a water quality
data website and automated data screening (Organics Detection System, 2011). Data from 1994-
2003 are available on ORSANCO's Organics Detection System web page as a Microsoft Access
database download.
96

-------
To further characterize spill incidents, ORSANCO developed a set of models to estimate travel
time and plume concentration. These models use the U.S. Army Corps of Engineers (USACE
FLOWSED model to calculate travel time and the USGS's Branched Lagrangian Model
Transport to estimate water quality (Gullick et al., 2003). Trained staff run the ODS sampling
programs as well as the travel time and water quality models, which can ideally be implemented
within one hour of notification of a spill (Gullick et al., 2003).
Communication and Response Network
In addition to coordinating sampling, travel-time modeling and other spill characterization
efforts, ORSANCO coordinates communications for spill incidents affecting the Ohio River.
Incidents can be reported from industries, river users or the National Response Center.
ORSANCO's communications system consists of an electronic bulletin board and direct phone
or fax communications with water utilities, water-dependent industries and state and federal
agencies (Gullick et al., 2003). The 2009 ODS upgrade added short message service (SMS) text
message, voicemail and email notification options for event detection notification (Organics
Detection System, 2011). An annual emergency response directory is published as a service to its
members and stakeholders.
Future Development
ORSANCO continues to plan improvements to its ODS operations to increase its ability to
detect, characterize, communicate and coordinate responses to major water quality incidents on
the Ohio River. Current ideas include integrating the spill model into a Geographic Information
System (GIS) platform, adding contaminant source data, links to material safety data sheets and
integration with health effects and treatment data. An emerging significant water quality
challenge for the Ohio River is cyanobacteria and ORSANCO has initiated efforts for addressing
this concern through monitoring and regional cooperation. Obstacles in that effort will be
extensive monitoring required for characterizing the watershed and Ohio River as well as the
need to develop predictive models.
Contact Information
Richard Harrison, Executive Director and Chief Engineer, ORSANCO, 513-231-7719 Ext. 105
Jason Heath, Director of Technical Programs, ORSANCO, 513-231-7719 Ext. 112
References
Brosnan, T. H. (1999). Early Warning Monitoring to Detect Hazardous Events in Water Supplies
(No. ILSI Risk Science Institute Workshop Report). International Life Sciences Institute
Press: Washington DC.
Gullick, R. W., Grayman, W., M., Deininger, R. A., & Males, R. M. (2003). Design of Early
Warning Monitoring Systems for Source Waters. J. Am. Water Works Assoc., 5(11), 58-
72.
Organics Detection System. (2011, March 22). Retrieved February 8, 2016, from
http://www.nexsens.com/case_studies/organics-detection-system.htm
ORSANCO. (n.d.). Ohio River Basin. Retrieved February 7, 2016, from
http://www.orsanco.org/ohio-river-basin
ORSANCO - Organics Detection System - ODS. (n.d.). Retrieved December 29, 2015, from
http://www.orsanco.org/organics-detection-system-ods
97

-------
Schulte, J. G. (2014, August). Early Warning Detection Training Workshop. Morgantown, WV.
Retrieved from http://www.horslevwitten.com/sourcewater/pdf/Schulte.pdf
98

-------
A 8. Susquehanna River Basin Case Study
Location
Susquehanna River Basin
Source Water
Coordinating
Agency
Start Date
Susquehanna River Watersheds
Susquehanna River Basin Commission (SRBC)
2003
System Description
Spanning 27,510 square miles in parts of Pennsylvania, New York and Maryland, the
Susquehanna River Basin is home to more than four million residents (SRBC, 2013a) (Appendix
Figure 12). The Susquehanna River runs 444 miles from Cooperstown, New York to the
Chesapeake Bay, and serves as a water supply for over 20 water systems in New York,
Pennsylvania and Maryland with an associated population of more than 2.5 million (Gullick et
al., 2004, 2006). Water in the region is a valuable resource not only for the provision of drinking
water, but also for agriculture, industry, energy development and recreation.
99

-------
MICHIGAN
WISCONSIN
IOWA
ILLINOIS
KANSAS MtSS0UR'
KENTUCKY
| Upper Mississippi River Basin	N
Slate Boundaries
100 200
MINNESOTA
Kilometers
INDIANA
"...A framework for innovative
partnerships and protocols for
fostering communication and
data sharing among water
suppliers, state/local agency
personnel, and the emergency
response community for the
purpose of enhancing drinking
water protection efforts"
(SRBM 2012).
Appendix Figure 12. Susquehanna River Basin EWS (SRBC, 2012).
The Susquehanna River Basin Commission (SRBC) led the development of the Susquehanna
River Basin EWS for the protection of water users dependent on the Susquehanna River.
Implemented in 2003, the EWS coverage area initially focused on the 12 Pennsylvania water
systems with intakes on the main stem of the Susquehanna River, since the system was largely
funded by the Pennsylvania Department of Environmental Protection (Gullick, 2004). With
support from New York State, the EWS was extended in 2006 to include the New York section
of the Susquehanna River Basin (SRBC, 2012). As the coordinating agency, the SRBC manages
the EWS, under the advisement and guidance of the water suppliers, environmental protection
agency personnel and emergency responders who comprise the stakeholder group. The current
coverage area of the EWS includes water suppliers serving approximately 700,000 people
(SRBC, 2013b).
The Susquehanna River Basin EWS consists of (1) a monitoring network, (2) a web and database
portal and (3) a communication and data sharing network.
100

-------
Monitoring Network
Water quality data are collected from nine monitoring points including water quality monitoring
of pH, temperature and turbidity with real-time data transmission. At four of the monitoring
points, TOC is monitored, while conductivity and DO are also measured at some locations
(SRBC, 2012). Monitoring points, located throughout the basin, are shown with green markers in
Appendix 11.
Data Management and Interpretation
The Susquehanna River Basin EWS database manages water quality monitoring data that is
accessible through a secure web portal. Online tools enable water suppliers to access and analyze
real-time water quality monitoring data. Features of the tools include a mapping tool for
visualizing real time water quality data and a time-of-travel tool for estimating contaminant
dispersal times. The web interface also provides discharger information and contact information
for emergency responders.
Communication and Response Network
The coupling of the water quality monitoring network with the SRBC's communication and data
sharing network enables access to the real-time monitoring data as well as important water
quality data collected by other agencies. For example, the PA DEP has monitoring stations along
the river. If an incident or irregular water quality is detected, the PA DEP notifies other water
users through the communication and data-sharing network. Water quality information is shared
between water suppliers in the same manner. The advanced notice of irregular water quality from
spills or other fluctuations (e.g., natural variation in water chemistry), enables timely adjustment
of operations at water treatment plants.
Future Development
In addition to the Susquehanna River Basin EWS for drinking water suppliers, the SRBC has
developed multiple additional programs to protect and manage Susquehanna River Basin water
resources including activities related to source water protection, low flow monitoring, flooding,
stormwater runoff and mine drainage, to name a few. The SRBC also implemented and manages
an EWS to detect potential contamination from natural gas drilling activities in the smaller rivers
and streams of the Susquehanna River Basin in parts of Pennsylvania and New York. Developed
in 2010, the Remote Water Quality Monitoring Network (RWQMN) consists of 59 monitoring
stations, with real-time measurements of pH, temperature, DO, conductivity and turbidity;
nutrients, metals and other constituents of interest are also measured four times per year (SRBC,
2015).
Challenges, Insights and Operational Notes
One challenge for the Susquehanna River Basin EWS is the need for a reliable and sustainable
funding structure to operate, maintain, expand and improve the EWS. While state agencies
assisted with start-up funding, the SRBC has been responsible for ongoing costs. Despite this
challenge, the success of the system can be attributed to systematic and gradual approach;
starting small with a limited number of stations and parameters, the system avoided over-
extension (Gullick, 2006).
101

-------
Contact Information
Susquehanna River Basin Commission, 4423 North Front Street, Harrisburg, PA 17110, (717)
238-0423, srbc@srbc.net.
References
Gullick, R. W., Gaffney, L. J., Crockett, C. S., Schulte, J., & Gavin, A. J. (2004). Developing
regional early warning systems for U.S. source waters. American Water Works Association.
Journal, 96(6), 68.
Gullick, R. W. (2006). Regional Early Warning Systems for Source Water Contamination-
PowerPoint Presentation - nj941d.pdf. Retrieved December 24, 2015, from
http://www.nj awwa.org/nj pdf/nj 941 d.pdf
SRBC, Susquehanna River Basin Commission. (2012). Information Sheet: Susquehanna River
Basin Early Warning System. Retrieved January 19, 2016, from
http://www.srbc.net/programs/docs/EWSInfoSheet020712.PDF
SRBC, Susquehanna River Basin Commission. (2013a). Information Sheet: Susquehanna River
Basin. Retrieved January 19, 2016, from
http://www.srbc.net/pubinfo/docs/SRB%20General%205_13%20Updated.pdf
SRBC, Susquehanna River Basin Commission. (2013b). State of the Susquehanna, 2013 Report.
Retrieved January 19, 2016, from
http ://www. srbc.net/ stateofsusq2013/docs/2013_SOTS_Report_Final_low_res.pdf
SRBC, Susquehanna River Basin Commission. (2015). Remote Water Quality Monitoring
Network, Data Report of Baseline Conditions for 2010-2013. Retrieved January 19, 2016,
from
http://mdw.srbc.net/remotewaterquality/assets/downloads/pdf/RWQMN_datareport_2010-
2013.PDF
102

-------
A 9. Upper Mississippi River Basin Case Study
Location
Upper Mississippi River Basin
Source Water
Coordinating
Agency
Start Date
Upper Mississippi River Watersheds
Upper Mississippi River Basin Association (UMRBA) and U.S. EPA,
Region 5
2003
System Description
The Upper Mississippi River (UMR) runs 1,300 miles from Lake Itasca in Minnesota to Cairo,
Illinois, with intakes for more than 20 public water suppliers along the river (UMRBA, 2016).
The UMR Basin spans approximately 189,000 square miles in parts of Minnesota, Wisconsin,
Iowa, Illinois and Missouri and is home to more than 30 million residents (Swanson, 2012)
(Appendix Figure 13). The UMR is a valuable resource for drinking water systems, as well as
power plants, industry, transportation, agriculture and recreation.
103

-------
MICHIGAN
WISCONSIN
IOWA
ILLINOIS
KANSAS MtSS0UR'
KENTUCKY
| Upper Mississippi River Basin	N
Slate Boundaries
100 200
MINNESOTA
Kilometers
INDIANA
"UMR-basedpublic water suppliers,
industries, and other partners have
supported efforts to establish an
"early warning monitoring network "
on the UMR which would serve to
provide advanced warning of a spill
event via continuous monitoring
installations" (IJMRBA, 2014).
Appendix Figure 13. Upper Mississippi River Basin (U.S. Fish and Wildlife, 2015).
Early development of the UMR Basin Early Warning Monitoring Network (EWMN) was
organized by American Water, a water utility in the region. With assistance from U.S. EPA
Region 5, further development was led by the Upper Mississippi River Basin Association
(UMRBA). UMRBA was developed as a joint effort between UMR Basin states for the
management of the water resources in the UMR Basin. UMRBA is comprised of governor-
appointed representatives from Illinois, Iowa, Minnesota, Missouri and Wisconsin (UMRBA,
2016). A pilot monitoring station was operated from 2003-2007 in Rock Island, Illinois. The
funding was provided by U.S. EPA and in-kind (non-monetary) assistance (UMRBA, 2007;
Swanson, 2012). Following the pilot project, the U.S. EPA became the primary coordinating
agency, with principal funding from the Regionally Applied Research Effort (RARE) grant.
Partnerships, collaborative efforts and the monitoring network were subsequently expanded. The
resulting collaboration includes the following partners: U.S. EPA, USACE, Shaw
104

-------
Environmental, other state and federal agencies, water utilities, the UMR Spills Group1,
UMRBA, universities and other stakeholders.
The UMRB EWMN consists of (1) a water quality monitoring network, (2) a web and database
portal and (3) a notification system.
Monitoring Network
The single station pilot system operated from 2003-2007 in Rock Island, Illinois, consisted of a
YSI Model 6600 Sonde sensor, which measured DO, chlorophyll, oxidation reduction potential,
pH, temperature, conductivity and turbidity. Online toxicity monitors were also piloted with
biomonitoring (Swanson, 2012).
The monitoring network includes six real-time monitoring stations that consists of a YSI probe, a
s::can Spectrolyser spectrometer, bi-valves and sensors, and a computer interface. The YSI probe
measures temperature, conductivity, DO, pH and turbidity. The s::can Spectrolyser spectrometer
measures turbidity, nitrate, TOC and DOC. Bi-valves equipped with sensors serve as online
toxicity monitors, with gape behavior used as a toxicity indicator.
Data Management and Interpretation
During the operation of the pilot system, the US ACE online data system, River Gages, was used
for data collection and management.
Communication and Response Network
Notification of adverse water quality is delivered through e-mail alerts, enabling timely
adjustment of operations at water treatment plants. In the discussion of biomonitoring, Allen et
al. (2014) presented a tiered response model; upon detection of a change in water quality, a
sample is collected, and a positive bioassay result leads chemical analysis followed by an
appropriate remedial/regulatory response.
Future Development
Goals of the UMRB EWMN include (Swanson, 2012):
•	Maintaining the existing monitoring network
•	Securing stable and sustainable funding
•	Addressing database needs
•	Considering improvements to the bio-monitoring algorithm
•	Maintaining and improve partnerships
Challenges, Insights, and Operational Notes
The pilot project highlighted the need for sustainable funding over the long-term as well as
challenges with respect to organizational coordination. With a multitude of stakeholders and a
wide-ranging coverage area, for the success of the UMR EWMN, a leading organization must act
as the coordinating agency (UMRBA, 2007). Similarly, Allen et al. (2014) stressed the
importance of collaboration given the project's scale.
1 The UMR Hazardous Spills Coordination Group includes members from US EPA, USCG, USACE, the U.S. Fish
and Wildlife Service, the Upper Mississippi River Basin Association (UMRBA), as well as environmental and
public health agencies in Illinois, Iowa, Minnesota, Missouri, and Wisconsin (UMRBA, 2014).
105

-------
Contact Information
Joel Allen, U.S. EPA/ORD/NRMRL, 26 W. MLK Drive, Cincinnati, OH 45268, (513) 487-
2806, allen.joel@epa.gov
References
Allen, J., Franz, W., Macke, D., and Panguluri. S. (2014). Source Water Monitoring and
Biomonitoring Systems. Retrieved January 20, 2016, from
http://www.horsleywitten.com/sourcewater/pdf/Allen.pdf
Swanson, G. (2012). A Unique Source Water Monitoring Initiative on the Upper Mississippi
River. Retrieved January 20, 2016, from
http://c.ymcdn.com/sites/www.isawwa.org/resource/resmgr/watercon2012-tuesday-
pdf/tuerespot830.pdf
UMRBA, Upper Mississippi River Basin Association. (2007). Pilot Monitoring Station at Lock
& Dam 15: Project Evaluation Report. Retrieved January 19, 2016, from
http://www.umrba.org/publications/hazspills/piloteval.pdf
UMRBA, Upper Mississippi River Basin Association. (2014). Upper Mississippi River Spill
Response Plan & Resource Manual. Retrieved January 20, 2016, from
http://www.umrba.org/hazspills/umrplan.pdf
UMRBA, Upper Mississippi River Basin Association. (2016). Upper Mississippi River Basin
Association. Retrieved January 19, 2016, from http://www.umrba.org/index.htm
U.S. Fish & Wildlife Service. (2015). Newsroom, Upper Mississippi River Basin Map. Retrieved
January 21, 2016, from http://www.fws.gov/midwest/news/343.html
106

-------
vvEPA
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
Office of Research and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300

-------