The Cadmus Group, Inc.
DRAFT
June 8, 1993
TECHNICAL GUIDANCE FOR ESTABLISHING
TOTAL MAXIMUM DAILY LOADS (TMDLs):
Involving CSO and Stormwater Point Sources and Nonpoint Sources
Project Officer
D. King Boynton
Prepared by
Jonathan B. Butcher, Ph.D.
J. Trevor Clements
Michael D. Marcus, Ph.D.
David Korn
The Cadmus Group, Inc., Durham, NC
Prepared for
U.S. Enviornmental Protection Agency
401 M Street, S.W.
Washington, D.C. 20460
-------
This Page Intentionally Blank
-------
The Cadmus Group, Inc. DRAFT June 10,1993
Table of Contents
Introduction iv
I. TMDL Problem Assessment, Goal Setting, and the Role of Models
1.1 What exactly is a TMDL? 1-1
1.2 The Water Quality Planning Context of the TMDL 1-6
1.3 The Relationship of TMDL Development to Watershed Protection .. 1-8
1.4 Phased TMDL Development and Basinwide Planning 1-10
1.5 Matching the TMDL Analytical Framework to Management Goals . 1-12
II. Use of Simple Models for TMDLs
2.1 Avoiding the Automatic Tendency Toward Complexity 2-1
2.2. Defining What is "Simple" 2-2
2.3 Function of Simple Models in the TMDL Process 2-2
2.4 Assessing Scoping Model Accuracy 2-5
2.5 Types of Simple Scoping Methods 2-7
HI. Simulation Models: Model Definitions and Available Models
3.1 Types of Models Used in TMDL Development 3-1
3.2 Introduction to Simulation Models 3-7
3.3 Characteristics of Available Simulation Models 3-16
IV. TMDL Modeling Strategy and Model Identification
4.1 Establishing a Modeling Strategy 4-1
4.2 General Criteria for a Modeling Strategy for TMDLs 4-5
4.3 Specific Model Issues for Simulation of Loading and Impacts from
Episodic, Wet-weather Events 4-16
4.4 Decision Criteria for TMDL Model Identification 4-23
V. Monitoring Plans and Impact Assessment for TMDLs
5.1 Interaction of Monitoring and Modeling for TMDLs 5-1
5.2 Existing Guidance for Information on Monitoring Designs and Sampling
Procedures 5-2
5.3 Defining General Monitoring Program Goals 5-6
5.4 Defining Specific Monitoring Program Objectives 5-13
ii
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
5.5 Specifying Sample Designs: Sample Sizes, Frequencies,
and Locations 5-14
5.6 Special Waterbody Considerations 5-24
5.7 Special Land-Use Considerations 5-39
VI. Sampling and Data Collection
6.1 Data Requirements 6-1
6.2 Assembling Existing Data 6-9
6.3 Sampling Techniques 6-17
6.4 Toxicological and Habitat Evaluation 6-46
6.5 Data Management Techniques 6-60
6.6 Quality Assurance and Quality Control - (QA/QC) 6-67
VH. Model Implementation, Interpretation and Analysis
7.1 Model Calibration and Validation 7-1
7.2 Model Accuracy and Reliability 7-11
7.3 Interpretation of Monitoring Data and Modeling Results 7-17
7.4 Evaluation of Effectiveness of Best Management Practices and Other
Control Strategies 7-22
7.5 Assessing Water Quality Impacts on Fish and Other Aquatic Life . 7-35
7.6 Interpretation of Biological Data 7-46
7.7 Physical Aquatic Habitat Evaluation 7-51
VIII. Establishing TMDLs/WLAs/LAsAdOS ,
in Caw
8.1 Which Comes First, the TMDL or WLAs/LAs? 8-1
8.2 Deciding When to Employ the Phased Approach 8-2
8.3 Splitting Up the Pie 8-4
8.4 Translating WLAs into NPDES Permit Requirements 8-9
8.5 Translating LAs into BMPs 8-11
8.6 Communicating the Results 8-12
IX. Case Studies
To be completed for Draft #2
iii
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
Introduction
The purpose of this document is to provide agencies responsible for developing
Total Maximum Daily Loads (TMDLs) with guidance on the identification, selection, and
use of technical frameworks or models which combine point and nonpoint, steady and
episodic source components. In April, 1991, the U.S. Environmental Protection Agency
(EPA) released a document entitled, "Guidance for Water Quality-based Decisions: The
TMDL Process" (EPA 440/4-91-001). That document provides broad outlines for several
aspects of the TMDL process including the regulatory basis, EPA policies and objectives,
methods overview, and agency responsibilities. The present guidance is intended to
augment that report by providing greater detail on actual methods of TMDL
development, with particular emphasis on how to model wet-weather point and
nonpoint source pollutant loading and resulting impacts within a given waterbody.
Historically, nonpoint source (NPS) pollutant runoff analyses and point source
(PS) waste load allocations (WLAs) have often been separated. WLAs have focussed on
municipal and industrial wastewater discharges, and only recently has greater attention
been given to control of wet-weather, episodic PS discharges, such as urban stormwater
and combined sewer overflows (CSOs). In fact, it is typical for the States and EPA to
have separate groups addressing wastewater discharges, episodic urban PS runoff, and
NFS runoff, and these separate groups may not interact on a frequent basis. With EPA's
recent emphasis on TMDL development and watershed protection, however, these
program functions are coming together. Agencies responsible for the the implementation
of die water quality program are quickly learning that successful water quality
management depends on an integrated PS and NPS control program. The challenge,
therefore, has become how to develop analytical frameworks that support combined
steady PS, episodic wet-weather PS, and NPS pollutant loading and impact analyses.
Integrating steady PS, wet-weather PS, and NPS analyses is not always a simple
task. Most of the commonly used WLA methods, as well as many State water quality
standards, are based on the concept of a relatively steady pollutant load. Steady loads
are of greatest concern when dilution flows are at their lowest point in the receiving
waterbody (i.e., during extended drought conditions), and are readily analyzed in terms
of design low flows, of a given probability of recurrence. In contrast, wet-weather loads
vary significantly in time and are typically greatest during high precipitation periods.
Determining how to combine these episodic loading events into a decision-making
framework originally based on design-flow analysis of PS impacts is often a formidable
task. Nonetheless, integration of episodic, wet-weather loads into the TMDL process is
a challenge that must be met to achieve an effective management strategy which protects
iv
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
the beneficial uses of impacted waterbodies. Accomplishing this integration will often
require the use of mathematical models to assess control strategies and predict the
frequency of water quality excursions. However, the presence of episodic, wet-weather
loads can lead to difficult and complex modeling problems.
This document provides guidance on the identification and selection of an
appropriate modeling strategy for estimating TMDLs involving episodic, wet-weather
loads. Such a strategy must recognize t ¦ physical complexity of many episodic loading
problems, yet also should recognize the practical constraints of time and money available
to complete modeling studies. How do we balance these competing needs?
Conceptually, the answer is straightforward: Use as simple a model as is appropriate to
calculate and apportion the TMDL. However, trying to pin down exacdy what
constitutes an appropriate level of complexity is a thorny problem; its resolution is a
principal aim of this document.
In addition to formulation of modeling strategy for estimating TMDLs, this
guidance provides: (1) a survey of simple and complex models for estimating wet-
weather loads and receiving water flows and impacts; (2) guidance on the development
of monitoring plans to support the TMDL estimation process; (3) information on data
collection and sampling for TMDLs involving wet-weather loads; (4) techniques for the
implementation and interpretation of simulation models; and (5) a selection of useful
case studies. The guidance is intended to be of general applicability to many types of
loading problems and receiving waterbodies, with extensive reference to other useful
EPA guidance.
v
-------
The Cadmus Group, Inc. DRAFT June 8,1993
Chapter I. TMDL Problem Assessment, Goal Setting, and
the Role of Models
Purpose: This chapter provides a brief overview of the concepts involved in
establishing a TMDL for watersheds with both point sources (PS) and nonpoint sources
(NPS). It sets the stage and provides the context for the main purpose of this document,
which is to provide guidance on the use of modeling to establish TMDLs which involve
episodic, wet-weather point and nonpoint source loads. In this context, wet-weather
point sources refer principally to permitted stormwater and combined sewer overflow
(CSO) systems. The regulatory context of TMDLs, which are designed to achieve
compliance with the requirements of the Clean Water Act (CWA), provides the basis for
assessing what questions should be answered in a modeling application.
1.1 What exactly is a TMDL?
The concept of Total Maximum Daily Load (TMDL) was established in 1972 with
the Clean Water Act (CWA) amendments to the Federal Water Pollution Control Act.
The reference to TMDL is found in Section 303(d) of the Act which states,
"Each State shall establish for die waters identified... las water quality limited].../
and in accordance with priority ranking, the total maximum daily load, for those
pollutants which the Administrator identifies under section 304(a) as suitable for
such calculation. Such load shall be established at a level necessary to implement
the applicable water quality standards with seasonal variations and a margin of
safety which takes into account any lack of knowledge concerning the relationship
between effluent limitations and water quality."
For the most part, however, implementation of this part of the act was delayed for
approximately 20 years as both EPA and States struggled to determine what exactly
constituted a TMDL.
In 1991, EPA took a giant step toward clarifying the meaning of TMDL with
publication of a guidance document titled, "Guidance for Water Quality Based Decisions:
The TMDL Process." Included among its contents are clarification of EPA's policies and
principals applicable under the CWA, the relationship of TMDL development to the
water quality planning and management process, and EPA/State responsibilities in the
development and implementation of TMDLs. Because of the comprehensive nature of
that document, it is not possible to recapitulate that material in full detail. Therefore,
the reader is encouraged to review that document as background material for
preparation in use of this guidance document. However, several of the main points will
be summarized below, and some will be revisited to provide further clarification on the
1-1
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
meaning of TMDL.
The statutory reference to TMDL in Section 303(d) has been interpreted by EPA
to mean that TMDLs are comprised of the sum of individual wasteload allocations
(WLAs) for point sources, and load allocations (LAs) for nonpoint sources and natural
background levels that are established for a given waterbody segment such that water
quality standards (WQSs) are maintained (U.S. EPA, 1991a). In addition, the TMDL
must include a margin of safety (MOS), either implicitly or explicity, that accounts for
the uncertainty between pollutant loads and the quality of the receiving waterbody.
Conceptually, this definition is denoted by the equation:
TMDL = I WLAs + 2 LAs + MOS
where: TMDL ->-4-4 maintains WQSs
Even with this clarification, however, several issues remain regarding what
practically constitutes a TMDL. Certain questions are routinely being asked by State and
EPA Regional personnel responsible for implementing TMDLs. Because these questions
require practical answers in order for TMDL development to move forward in a
consistent and orderly fashion, they are addressed below item by item:
Question: Is the TMDL comprised of all parameter restrictions that collectively
protect water quality, or is there a separate TMDL for each water quality
parameter?
Answer: EPA recommends that there be one TMDL per parameter for a defined
waterbody management unit. While - under this approach - several
TMDLs may be required to protect water quality standards, practical
considerations necessitate that control strategies address specific
parameters. It is possible, however, that several types of parameters will
have to be controlled in order to address a single water quality standard.
In these cases, the administering agency should make its own
determination on the most effective and efficient way to establish and track
the TMDL (i.e., as separate or collective requirements).
Question: Do TMDLs only apply to parameters for which numeric criteria have been
adopted?
Answer No, TMDLs are intended to address all State water quality standards
including narrative criteria, numeric criteria, designated uses, and
antidegradation provisions. Thus, parameters such as physical habitat and
biological integrity are included.
1-2
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
Question: Are TMDLs always to reflect a single numeric value which constitutes the
total loading capacity for the given waterbody?
Answer: No, it is not realistic to think that a single total loading restriction will
always be definable for a given waterbody. The loading capacity (LC) or
assimilative capacity (i.e., the greatest amount of loading that a waterbody
can receive without violating water quality standards - 40 CFR 130.2(g)) of
a waterbody does not necessarily reflect a fixed amount of loading. In this
regard, the "pie" diagram commonly used to explain TMDL components
(see figure 1-1) may actually be misleading because of its simplistic
configuration. One tends to think of a pie as a fixed object that can be
"sliced" or allocated into readily measurable parts. However, assimilative
capacity often varies in time and space due to the dynamic, sometimes
random nature of the ecological features (e.g., physical, chemical, and
biological) which comprise a waterbody and its watershed. Thus, rather
than being represented by a single numeric value, a TMDL is often
comprised of a combination of management strategies for a given pollutant
that collectively protect water quality standards. Those strategies may
include seasonal or multi-level controls to address this variation.
In this manner, the TMDL at the mouth of a watershed is not simply the
critical flow at die mouth of the watershed multiplied by the ambient
criteria. Such a definition would ignore the assimilative properties (e.g.,
oxidation, volatilization, nitrification, sediment adsorption, etc.) of the
watershed above the mouth. It is quite possible to have a much greater
load distributed throughout a watershed that will not result in a violation
of water quality standards than would be allowed at a fixed point at the
mouth of the watershed. Alternatively, one could conceive of situations
where allowable loading could increase as long as instream concentrations
are maintained below levels necessary to protect the standard (e.g., cases
where wasteflows dominate streamflow and increase without increases in
pollutant concentration). Thus, the TMDL may be thought of as "a tool for
implementing State water quality standards [which] is based on the
relationship between pollution sources and in-stream water quality
conditions. The TMDL establishes the allowable loadings or other
quantifiable parameters for a water body and thereby provides the basis
for States to establish water quality-based controls. These controls should
provide the pollution reductions necessary for a waterbody to meet water
quality standards." (U.S. EPA, 1991a).
1-3
-------
LAs for Nonpoint Sources
and Natural Background
Margin of Safety
WLAs for Point Sources
Figure 1-1. Allocation of Waterbody Assimilative Capacity
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
Question: Is there a specific management unit (i.e., size of watershed) for which the
TMDL must be established?
Answer Management units should generally be chosen to fit the extent of the
problem being addressed. Therefore, there are no regulatory requirements
for choice of waterbody segments. There are, however, a few conventions
that make TMDL development more technically and scientifically sound.
In general, management segments should be extended upward within a
watershed to the basin boundaries. Since TMDLs include nonpoint source
components, use of the watershed as a management unit ensures that all
potential contributors can be addressed. While an isolated portion of the
watershed may be the only segment exhibiting impairment, contributions
to that impairment may be coining from upstream sources.
As with any rule, there are exceptions to it. There may be natural
breakpoints (e.g., lakes or impoundments, etc.) or the system may be so
large that it is more practical to break it into segments. EPA has been
recommending that States track the TMDLs using the Waterbody System,
so the management unit that corresponds to what States input to that
system may be a logical choice for those that use it
It may also be helpful to develop and track TMDLs according to
hydrologic units that correspond with other agencies dealing with water
such as the USGS and SCS. Even if management units are not exactly the
same size between coordinating agencies, sharing and interpretation of
information can be aided if the management units "nest" within each other
so that information can be aggregated and compared at some level. For
example, the State of Virginia uses SCS hydrologic units which have been
nested within the USGS cataloging hydrologic units. This can be very
important where tools such as geographic information systems (G1S) are
used to manage, interpret, and present watershed information.
Question: Should the margin of safety (MOS) term always be made explicit and does
it also include a reserve for future allocations?
Answer The TMDL Process document. (U.S. EPA, 1991a) indicates that the MOS is
typically incorporated with conservative assumptions in calculations or
models used to develop TMDLs (i.e., implicitly). This implies that the
estimate is biased on the side of safety. However, the premise that
assumptions actually provide conservative estimates may need to be
scrutinized on a case-by-case basis. As • the EPA Process document
indicates, where an additional MOS is needed, an explicit MOS term can
1-5
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
be added to the equation.
It should be recognized, however, that in some cases the TMDL developer
may not be able to establish an MOS to completely account for uncertainty.
This can occur when the uncertainty is so large that it becomes impractical
to remove that large of a portion of loading. It also can occur when
assimilative capacity has been exhausted and existing sources are already
at state-of-the-art control levels. In these cases, agencies are often
struggling just to find additional ways to reduce pollution sources to get
near targeted reduction levels, and there is no additional capacity to hold
in reserve to account for uncertainty.
On the other hand, the TMDL developer may run into cases where the
sum of WLAs and LAs under current conditions leaves a portion of the
assimilative capacity remaining. This is fine as long as the TMDL reflects
control strategies that collectively keep sources of pollution at or below the
loading capacity. In fact, this may be done intentionally to allow for future
uses and thus the TMDL becomes an effective long-term planning tool.
Translating TMDL strategies into LAs and WLAs will be addressed further
in chapter 7.
1.2 The Water Quality Planning Context of the TMDL
Prior to beginning the TMDL development process, if s important to understand
the overall planning context in which the TMDL development process fits. The TMDL
developer does not act in isolation from the rest of the water quality management
program for a given State. Rather, TMDL development occurs in response to water
quality assessment and subsequent prioritization of a waterbody for management action.
This process is known as a water quality-based approach to pollution control and is well
outlined in chapter 2 of the EPA manual, "Guidance for Water Quality-based Decisions:
The TMDL Process." The approach can be simplified into the following five step process
(illustrated in figure 1-2):
Step 1: Identification of Water Quality-Limited Waters
Step 2: Priority Ranking and Targeting
Step 3: TMDL Development
Step 4: Implementation of Control Actions
Step 5: Assessment of Water Quality-Based Control Actions
Section 303(d) of the Clean Water Act (CWA) requires States to establish TMDLs
for waterbodies where federally-based minimum guidelines for wastewater discharges
1-6
-------
1.
Identification of Water Quality-Limited
Waters
. Review water quality standards
• Determine if adequate controls are
in place
5.
Assessment of Water Quality-Based
Control Actions
• Monitor point/nonpoint sources
• Evaluate TMDL for attainment of water
quality standards
2.
Priority Ranking and Targeting
• Integrate priority ranking with other
water quality planning and management
activities
Use priority ranking to target
wsterbodies for TMDLs
4.
Implementation of Control Actions
. Update water quality management plan
. Issue water quality-based permits
. Implement nonpoint source controls
(section 319 management plans)
3.
Development of TMDLs
• Apply geographic approach where
applicable
• Establish schedule for phased
approach, if necessary
• Complete TMDL development
Figure 1-2. General Elements of the Water Quality-Based
Approach (from U.S. EPA, 1991a)
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
are not stringent enough to protect water quality standards. Through the 303(d) process,
States must identify and prioritize those waters in need of TMDL management strategies.
Thus, to a large extent, water quality problems have already been subject to a
preliminary assessment and the major causes and sources of impairment have been
identified by the time at which the control agency is ready to establish a TMDL for a
given waterbody.
The technical core of the process ; contained in Step 3. In this step, the TMDL
developer begins with the identification of a prioritized, targeted waterbody and
develops an analysis sufficient to develop appropriate control actions. It is in this area
that modeling typically plays an important role, and subsequent chapters can be thought
of as providing technical guidance in support of the completion of Step 3 of the TMDL
development process for TMDLs involving wet-weather point and nonpoint sources.
1.3 The Relationship of TMDL Development to Watershed Protection
The method that EPA has chosen to implement the Section 303 planning process
recognizes that States cannot address all identified water-quality limited waterbodies
simultaneously. After recent revisions (see 57 Federal Register 33040; Friday, July 24,
1992), the regulations implementing Section 303(d) now only require that States identify
the waterbodies that they have prioritized for TMDL development over a two year
period. A critical part of the TMDL process is realistically determining which
waterbodies can be handled at a given time and establishing a logical schedule for
TMDL development.
EPA also recommends that development of TMDLs proceed along a geographic
basis (i.e., watershed), since water quality concerns are usually area wide and caused by
multiple sources (U.S. EPA, 1991a). Some States (e.g., NC and SC) have established their
TMDL development schedule around a five-year basin wide planning cycle.
Representatives of these States have indicated that using the "basin" as the unit of
management has provided for greater efficiency, thereby allowing the development of
a larger number of more comprehensive TMDLs in a given amount of time (U.S. EPA,
1992). Elements of the basinwide planning process include comprehensive water quality
monitoring, assessment, waterbody management prioritization and modeling analyses,
all of which combine to provide a ready-made foundation for TMDL development.
These activities are scheduled over the five-year cycle such that by its end, all TMDLs
in the basin are implemented via an approved basinwide management plan. Thus,
TMDL problem assessment and goals are established within the watershed management
planning process. This allows the TMDL developer to see the "big picture" of TMDL
needs within the basin and consequently provides better background information to
make critical decisions regarding the level of complexity and resources that will be
1-8
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
devoted to a given TMDL.
For example, in North Carolina a basin is re-examined every five years and the
basinwide management plan is updated accordingly. During the first part of the five-
year period, the State identifies known problem areas and establishes intensive field
sampling plans for model development. Both ambient and intensive monitoring occurs
during years one through three of the five year period for that basin, and model
development begins as information is collected. Model development and application
should be completed during year four and TMDL results should be incorporated into
the plan that goes through the public review and adoption process in year five.
Therefore, those prioritized waters that did not receive a TMDL in one planning cycle
can be targeted for completion within the next cycle. This occurred with their recent
adoption of a plan for the Neuse River Basin.
The State of New Jersey recently proposed a rule that would formally adopt a
TMDL approach within a basinwide planning context which specifically addresses
modeling concerns in a cyclic review format. After dividing the State into five
geographic regions and establishing a schedule for water quality program activities for
each region, New Jersey's action plan includes the following TMDL development
components for each watershed:
1. Select the parameters needing a water quality model for each watershed
in a given permit cycle.
2. Determine the appropriate model complexity and accuracy for the TMDL
for each parameter that needs to be addressed.
3. Determine the priority for completing the TMDL analysis for each
parameter for each watershed.
4. Complete the TMDL for each parameter of interest.
Similar to North Carolina's program, the New Jersey approach would handle TMDL
development within a five-year cycle:
Year 1: Preliminary instream data for all toxics are gathered at selected sampling
points within die watershed. These data will be used to assist in
determining which parameters might be targeted during the first permit
cycle. All other available instream data are collated. Effluent data
collected as a result of the monitoring requirements in the discharge
permits are collated. A list of the potential parameters to be targeted is
selected from this limited database. Preliminary evaluation may be
1-9
-------
The Cadmus Group, Inc. DRAFT June 8, 1993
undertaken with scoping models at this stage.
Year 2: Instream and effluent data are collected for the targeted parameters. If
conventional and nonconventional parameters have not yet been addressed
in the watershed, or in portions of the watershed, those parameters will be
addressed in the sampling and subsequent model development.
Y ear 3: The hydrodynamic water quality model is developed focusing on selected
parameters. Preliminary wasteload allocations for point sources and load
allocations for nonpoint sources are developed for the selected parameters.
Year 4: The public input process for the selected parameters is completed.
Discharge permits are drafted at the end of TMDL/WLA/LA public input
process.
Year 5: By the end of Year 5, the affected permits in the watershed are issued as
final permits.
During a given 5-year cycle, not all parameters in each watershed can be
addressed. The objective is to begin with those parameters which have the highest
priority. Lower priority parameters can be addressed in subsequent cycles.
1.4 Phased TMDL Development and Basin wide Planning
Problem assessment and goal setting for TMDLs can be limited by lack of
information on sources and loads. In particular, accurate determination of nonpoint
source loads is often difficult. However, EPA regulations (40 CFR 130.2(g)) provide that
load allocations for nonpoint sources "are best estimates of die loading which may range
from reasonably accurate estimates to gross allotments..." Thus, an incomplete
understanding of nonpoint source loading to an impaired waterbody should not delay
the implementation of water quality-based control measures.
To address the development of TMDLs in cases where estimates are based on
limited information, EPA (1991a) has suggested use of a phased approach. The phased
approach is defined "as a TMDL that includes monitoring requirements and a schedule
for re-assessing TMDL allocations to ensure attainment of water quality standards." EPA
guidance also states that the phased approach may be "necessary" where nonpoint source
controls are involved. In order to allocate loads among both nonpoint and point sources,
there must be reasonable assurances that nonpoint source reduction will in fact be
achieved. However, the only federally enforceable controls under the CWA are those
for point sources through the NPDES permitting process. With the phased approach,
1-10
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
the TMDL includes a description of the implementation mechanisms and the schedule
for the implementation of nonpoint source control measures.
1.4.1 Phasing Management Actions
The phased approach allows for an incremental approach to managing water-
quality limited waters. Control agencies can either implement interim measures while
a final TMDL strategy is being developed or they can hold off on implementing a
strategy until the appropriate information is collected. In either case, however, the
agency should develop a firm schedule for collection of needed information and final
development of the TMDL management strategy.
1.4.2 Accomodating Long-Term Modeling Studies
A phased approach may be a good choice where extensive model development
is required to provide TMDL developers with a mechanism for predicting the outcomes
of alternative management strategies. Complex modeling typically requires a
considerable amount of time for data collection, model calibration, and model validation.
Therefore, a phased approach can allow the developer time to establish a coherent
modeling strategy. Simple models can be used for initial scoping of the problem and
preliminary assignment of WLAs and LAs. Often, earlier phases of data review and
analysis focus on establishing relative loadings and therefore require less precision. The
scoping exercise can lead to the formulation of a focused sampling plan, which in turn
provides the basis for more detailed modeling in later phases of the process where a
higher degree of precision may be warranted.
1.4.3 Piggy-backing With Basin wide Planning
The phased approach works well within a basin wide planning context, in which
a regular cycle of review of TMDLs is established by scheduling concurrent review of
all NPDES permits within a basin. Phased TMDLs are naturally complementary to
regular basinwide TMDL reviews, and the timing of the phases can be timed to the
overall review process. In North Carolina, the phased TMDL has been integrated with
the five-year review basinwide review cycle. Of the more than fifty TMDL strategies
outlined in North Carolina's 1993 plan for the Neuse River Basin, several were
implemented in a phased format. For instance, while a TMDL strategy was adopted for
the upper 180 mile stretch of the main stem of the river using a field calibrated QUAL2E
model, an interim strategy was adopted for the lower estuary portion of the basin while
a more complex, multi-dimensional, hydrodynamic model is being developed. The State
will include any modifications to the Neuse Basin plan in 1998 as it comes up for
1-11
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
renewal.
1.5 Matching the TMDL Analytical Framework to Management Goals
Choosing the right analytical framework (e.g., model or combination of models)
for TMDL development requires the developer to consider factors that match
management goals. The goals can range "om general water quality program objectives
to meeting specific criteria within the waterbody of concern. In addition, the developer
must evaluate methods to achieve the goals in light of resource constraints. The
following details some of the factors that will affect the decision regarding model choice.
1.5.1 .Meeting Water Quality Standards
The overall goal of TMDL development is to achieve compliance with water
quality standards (WQSs), including narrative criteria and designated uses. Modeling
requirements for the TMDL are necessarily constrained by the need to provide answers
that address the excursions of WQSs, and thus depend on the form in which these WQSs
are expressed (i.e., narrative versus numerical; frequency of excursion allowed; etc.).
TMDLs can address not only specific chemical concentration requirements, but
other types of WQSs as well. As stated in EPA's Guidance for Water Quality-based
Decisions...,"..it is becoming increasingly apparent that in some situations water quality
standards - particularly designated uses and biocriteria can only be attained if non-
chemical factors such as hydrology, channel morphology, and habitat are also addressed.
EPA recognizes that it is appropriate to use the TMDL process to establish control
measures for quantifiable non-chemical parameters that are preventing the attainment
of water quality standards..."
In some cases, this means controlling parameters for which there are no specific
standards. For instance, States rarely adopt instream phosphorus standards since the
nutrient by itself is not a threat to human health or aquatic life and habitat. Rather, it
is the eutrophication problems (e.g., substantial fluctuations in dissolved oxygen,
excessive algae growth, fish kills, etc.) that are associated with excessive amounts of the
nutrient being loaded to the system. The standards tend to reflect the response
variables. However, as relationships are formalized between the response and
dependent variables, TMDLs can be established for parameters for which no formal
criteria have been adopted.
The statistical form of the standard should also be considered in making a
decision on the appropriate model framework. Do we need to calculate a long-term
average, a "typical" concentration at design flow, or actually evaluate the probability of
1-12
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
water quality excursions? EPA recommends that water quality criteria statements
Should be developed in a duration-frequency format, to include requirements that a
given concentration not exceed a critical value on average more than once in a given
return period. As described in U.S. EPA (1991b), EPA criteria are developed as national
recommendations to assist States in developing their standards and to assist in
interpreting narrative standards. EPA criteria or guidance consist of three components:
• Magnitude: How much of a pollutant (or pollutant parameter such as
toxicity), expressed as a concentration, is allowable.
• Duration: The period time (averaging period) over which the instream
concentration is averaged for comparison with criteria concentrations. This
specification limits the duration of concentrations above the criteria.
• Frequency: How often criteria can be exceeded.
(Note: While these components are often thought of in regard to numeric standards, its
important to point out that the concepts of magnitude, duration, and frequency can be
applied to non-chemical water quality problems as well.)
When expressed in this form, a typical aquatic life water quality criteria statement
is recommended to contain a concentration, averaging period, and return frequency. For
instance, U.S. EPA (1985) recommends that water quality criteria for the protection of
aquatic organisms should typically be expressed in a generic form as follows: "...aquatic
organisms and their uses should not be affected unacceptably if the four-day average
concentration of [the pollutant] does not exceed [the lower of the chronic-effect or
residue-based concentrations as the criterion continuous concentration] more than once
every three years on the average and if the one-hour average concentration does not
exceed [the acute effect-based criterion maximum concentration] more than once every
three years on the average."
A duration-frequency criteria statement directly addresses protection of the
waterbody; that is, it is expressed in terms of the acceptable likelihood of excursions of
WQSs. While this appears ideal, it may not always be prafctical, as it requires estimation
of long-term averages. Many states rely instead on the older concept of critical flows
(e.g., 7Q10). Setting a standard based on. a given critical Jlow can be shown to be
equivalent to a specification of an average duration and frequency of excursion, at least
for a load which is independent of flow. However, the concept of a single critical flow
may not make sense when dealing with wet-weather episodic loads, as these are likely
to be correlated with flow; thus the high loads are very unlikely to occur simultaneously
with critical low-flow conditions. The critical flow based LA may then be overly
conservative for parameters with a long response time, but might not be protective
1-13
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
enough for a nonconservative pollutant, such as a toxic with a short response time,
where critical concentrations are achieved not at low flow but during storm runoff
events. A more accurate analysis would require predicting the actual duration and
frequency of WQS excursions.
The modeling analysis will generally need to provide answers which match the
form in which WQ criteria and permit conditions are expressed. If criteria and permit
conditions take, or are recommended to take a duration-frequency form, then modeling
for TMDLs should provide similar information: that is, the modeling and monitoring
activities should eventually result in an estimate of the frequency distribution of
receiving water concentrations, rather than just worst-case and/or average estimates.
This is particularly important for episodic, wet-weather loading, as the concentrations
and impacts of these types of sources are sensitive to variability in both runoff rates and
streamflow. However, worst-case estimates (which imply use of a large MOS) may be
useful in the early stages of phased TMDL development due to their relative simplicity.
1.5.2 The Role of Models in the TMDL Development Process
The link between WQSs and pollutant loads is usually provided by a combination
of modeling and monitoring. In addition to addressing the WQSs, the TMDL developer
must consider several other factors including the amount of information available on
which to base the decision, the complexity of the problem, spatial and temporal
resolution, and the availability of program resources for TMDL development and
implementation. The first issue1 that the TMDL developer faces after receiving a
prioritized waterbody is whether enough information already exists to develop a TMDL
and whether a modeling analysis is needed to help establish that TMDL. The answer
to this question will determine the future level of effort that needs to be employed and
may lead to a series of decisions regarding model choice and TMDL development (see
Figure 1-3).
Figure 1-3 illustrates the general flow of the TMDL development process, with
particular emphasis on the decision-making junctures where model development is
required. The remainder of this guidance document will focus on components of this
decision-making framework and how they apply to TMDL development for problems
involving wet weather point and nonpoint sources.
1-14
-------
WQSs achieved
~
WQSs not
achieved
Problem waterbodies, parameters of
concern, and causes/sources provided
from assessment (i.e., 305(b), 304(1),
314(a), 319(a) and prioritization (i.e.,
^ 303(d)) processes
Figure 1-3. Flow Chart of the TMDL Development Process
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
1.5.3 Key Issues in Modeling for TMDLs
What sort of models should be used to establish a TMDL? The focus of the
remainder of this guidance is the identification and application of models for calculating
TMDLs which involve episodic, wet-weather loads. Key issues to be addressed in
subsequent chapters include the following;
• How should episodic, wet-v ither PS and NPS be represented in modeling for
TMDLS? TMDLs including wet weather point source and NPS necessarily
involve loadings which are episodic, or unsteady, dependent on die pattern
of rainfall, and which must be described statistically. Further, the TMDL
will often address the interaction of PS and NPS on a watershed basis.
Detailed simulation of PS and NPS loading in response to precipitation
events can require extensive work with complex models. However,
simpler approximations may also be available. What level of
representation of the wet-weather PS and NPS loads is appropriate in
terms of protection of the receiving waterbody?
• What is the appropriate level of model complexity? When is a complex model
needed? And when is a simpler modeling approach appropriate? The
questions are simple, but the answers are not "Complexity" may refer to
the hydrodynamic or pollutant transport component of models. These
components do not necessarily need to be at the same level of complexity.
Similarly, it may be appropriate to model NPS and receiving waters at
different levels of complexity. Although wet-weather sources may be at
issue, perhaps only relative long-term comparisons are needed and thus
the complexity of die mechanisms can be ignored. In addition, increasing
model complexity does not necessarily increase predictive accuracy. In
general, we expect project costs to increase as model complexity increases.
How do we strike the appropriate balance between complexity, accuracy
and cost?
9 When are simple models appropriate? It is clear that simple models have a
definite place at the scoping level of most problems. That is, before
sufficient data are available to undertake a complex model, approximate
answers can be obtained using simpler approaches, as described in Chapter
n. The results can provide the basis for the development of a monitoring
plan, as well as the development of a more complex model of the
waterbody. However, there will also be cases where it may be
appropriate, or necessary, to calculate the TMDL using simple models
alone. This may be adequate if (1) the uncertainty introduced into the
analysis by use of the simplifying assumptions can be and is evaluated (at
1-16
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
least qualitatively) arid is incorporated into the margin of safety, and (2)
the spatial resolution of the model is adequate to address the appropriate
WQSs.
When are complex simulation models required? Some States have proposed
that complex hydrodynamic models will be required to establish TMDLs
in most situations involving wet-weather PS and NFS loads. Such
situations may require use of complex models to achieve an accurate
representation of response to specific precipitation events. However, it
may be possible to meet the goals of the CWA and comply with the
intentions of the TMDL process by using simpler, approximate models.
The simpler models are attractive on the basis of lower cost of
implementation, but may misrepresent the actual response of the
waterbody. If resources are not available to apply complex models to all
TMDLs, what are the conditions in which use of complex models is most
essential to provide protection of the waterbody designated uses?
References, Chapter I
U.S. EPA. 1992. Watershed Cost Survey. Assessment and Watershed Protection
Division.
U.S. EPA. 1991a. Guidance for Water Quality-based Decisions: The TMDL Process. EPA
440/4-91-001. Assessment and Watershed Protection Division.
U.S. EPA. 1991b. Technical Support Document for Water Quality-based Toxics Control.
EPA/505/2-90-001. OWEP/OWRS.
U.S. EPA. 1985. Guidelines for Deriving Numerical National Water Quality Criteria for
the Protection of Aquatic Organisms. NTIS PB85-227049.
1-17
-------
The Cadmus Group, Inc.
DRAFT
June 4, 1993
Chapter II. Use of Simple Models for TMDLs
Purpose: Chapter II will provide more information on how simple scoping
methods can be effectively used for assessment and TMDL development, particularly
where PS and wet-weather NPS pollutant loading must be considered jointly in the
modeling analysis. Functions of simple models are discussed, along with factors to
consider in the trade-off between model accuracy and the cost of implementation.
2.1 Avoiding the Automatic Tendency Toward Complexity
Given the complexities of the transport and ecological processes in most
waterbodies, the modeler may be tempted to think that a complex model is required to
capture the behavior of the system. However, it is important to recognize that capturing
complexity may come at a high price. It also may not be necessary to address the issue
at hand, which is compliance with WQSs and protection of designated uses. For
example, although the real time transport of nutrients from various land use activities,
stormwater runoff, and point sources in a lake's watershed may be complex, capturing
that complexity may not be relevant to a TMDL if the lake's eutrophic response is more
a function of average total seasonal or total annual loading. As a general rule of thumb,
simple models are preferred when they satisfy the objectives of the user, because they
are usually developed at a lower cost and their basis is typically better understood by
decision-makers and stakeholders. Therefore, simple scoping models should not be
ignored or bypassed in favor of complex simulation models without an assessment of
whether a simple model can provide an adequate answer.
There are many factors that a modeler can consider in answering the question of
how complex a model is needed. Chapter IV will cover these factors in greater detail,
but in general the factors include technical, regulatory, and user criteria. While it may
be a challenge to find the right balance among these factors, it may help die modeler to
keep a few modeling fundamentals in mind when going through this decision process
(Donigian and Huber, 1991):
Have a dear statement of project objectives. Verify the need for quality
modeling.
Use the simplest model(s) that will satisfy the project objectives.
To the extent possible, utilize a quality prediction method consistent with
available data.
1.
2.
3.
2-1
-------
The Cadmus Group, Inc.
DRAFT
June 4,1993
4 Only predict the quality parameters of interest and only over a suitable
time scale (i.e., don't attempt intra-storm variations in quality unless it is
necessary).
2.2 Defining What is "Simple"
There is no explicit definition of what constitutes a simple model; rather, the
definition must be made in relative terms. Any model is a simplification of reality.
However, some models are more detailed and complex than others in their
representation of the ways in which a watershed's processes vary in time and space. A
"simple" model can be defined as one in which the finer scale details, although known
to exist, are intentionally overlooked. For example, an approximate empirical prediction
of an NPS event mean concentration (EMC) made based on average land use
characteristics, without detailed simulation of the underlying processes, is obviously a
simple approach.
In general, we can define "simple" models as approximative scoping tools that are
relatively easy to implement and which provide an order of magnitude (or better)
assessment of a water quality problem. They are characterized by relatively large spatial
scales, and large temporal scales, often steady-state. Analytical solutions to transport
equations provide a clear example of simple models that can be used for approximate
scoping of water quality problems. Analytical models generally assume rather restrictive
conditions, typically one-dimensional, constant parameters and steady state. While real
world conditions may not meet any of these criteria, judicious application of analytical
solutions is very useful to "home in" on the range of potential responses and relative
importance of various phenomena.
Simple models may also utilize computer codes with numerical solutions, but
implemented at a relatively coarse spatial and temporal scale; While there is no clear
line between simple and complex, our intention is to show the practical uses of models
which impose a relatively low burden of data collection and model calibration and treat
dynamic phenomena in a maimer which is highly averaged in time and space.
2.3 Function of Simple Models in the TMDL Process
Simple models can play a significant role in three areas of the TMDL development
process: framing the problem, identifying and prioritizing problem areas, and in actually
establishing LAs and WLAs. The first two uses are obvious; the third is somewhat more
controversial.
2-2
-------
The Cadmus Group, Inc.
DRAFT
June 4, 1993
2.3.1 Framing the Problem
Sometimes, when there are limited data, the modeler feels at a loss as to which
step to take next. Simple models can often help in these situations by "framing" the
problem. For instance, a lack of data may leave the modeler wondering about the
magnitude of the problem. Use of a simple model (e.g., USGS regression model or
FHWA model) may give the modeler a better approximation of the problem's
magnitude, and perhaps place an upper nd lower bound to it.
In addition, the simple model application may provide a guide for future
activities. Insights may be gained for development of a monitoring program such as key
sample locations or parameters that warrant closer attention than others. Future
modeling activities may be guided as well. For instance, evaluation of the results of the
simple model application may show that a more complex model is needed to accurately
address the problem. By exposing information or conceptual model gaps, the modeler
is left with a better understanding of what needs to be done to achieve final objectives.
2.3,2
Prioritizing
Targeting and
Simple models are often an
excellent choice where the objectives
are to identify and prioritize
waterbodies or pollutant sources of
concern. In these cases, the modeler is
typically in need of only gross
characterization for relative
comparison. By comparing
approximate loads between sources or
estimated conditions among
waterbodies, the modeler can target
those areas of most interest or
importance regarding further
investigation or management.
New Watershed Screening Tool
EPA/QST recently released a
prototype of a tool for identifying and
targeting problem waterbodies titled the
"Watershed Screening & Targeting Tool
(WSTrr WSTT summarizes STORET
water quality and flow data for a
watershed. It operates in a user-friendly
PC environment, and allows the user to
screen, compare, and target sections within
a particular watershed to locate pollution
hot spots that need special attention, it
also features a watershed model that
predict daily runoff, streamflow, erosion,
sediment load, and nutrient washoff.
While currently only available for the
States of Alabama and Georgia within
Region IV, EPA is considering expanding
the application of the WSTT to other
regions of the country.
Creation of a nutrient budget is
a good example where simple scoping
models provide these functions.
Based upon a rudimentary breakdown
of land use within a given watershed,
the modeler can use export coefficients to estimate annual nutrient loading from each
2-3
-------
The Cadmus Group, Inc.
DRAFT
June 4,1993
land use category (i.e., unit loads such as kg/hectare/yr). These data, combined with
PS data collected through the NPDES self-monitoring program, can be used by the
modeler to set up a budget table that provides an estimate of total loading and allows
for comparison of source loads. Such comparisons can show which sources may be of
greatest concern on a long-term average basis, and the modeler can determine whether
a management decision should be based on that limited information. In cases where the
implications of the decision are potentially far-reaching or costly, the modeler may
choose to further investigate the problem focusing on those sources identified as most
important via the screening process. Similarly, this process could be repeated for a
number of smaller hydrologic units within a larger basin such that relative comparisons
can be made between those watersheds, and efforts or decisions could be prioritized
accordingly.
2.33 Establishing LAs/WLAs
Finally, simple models can also be used to establish TMDLs/LAs/WLAs under
certain conditions. After evaluating appropriate factors (see Chapter IV), the modeler
may select the simple model(s) as the best choice(s) to use for development of LAs and
WLAs within the TMDL management strategy. This would imply that the reliability of
the model predictions has been deemed adequate for management purposes. This can
be true even where the MOS is relatively large, particularly where the outcome or
response to the model result would not change substantially within the range of error
in Hie prediction. For example, perhaps the decision to target a portion of a watershed
for stormwater controls is simply based on the likelihood that the loading from
uncontrolled sources would exceed a given threshold. If the model results show that the
loads far exceed the threshold, even after considering a large potential model error, then
a more accurate model is not needed to justify the application of controls.
It is well established that simple models may often suffice for TMDLs for waters
whose impacts are largely or entirely due to point sources. Traditionally, EPA and
States have focused on the control of point sources via WLAs written for an individual
discharger, rather than from a whole basin perspective. Under these conditions,
relatively simple steady-state analytical models are often sufficient to evaluate WLAs.
Use of such models is well covered in EPA's series of technical guidance manuals for
performing wasteload allocations, and these methods may continue to suffice in areas
where point sources are the main concern.
TMDLs which address wet weather NPS, on the other hand, necessarily involve
loadings which are episodic or unsteady - dependent on the pattern of rainfall - and
which must be described in terms of a statistical probability of recurrence. Further, the
TMDL concept typically includes examination of the interaction of PS and NPS on a
watershed basis. Performance of an accurate TMDL (i.e., one that includes a minimal
2-4
-------
The Cadmus Group, Inc.
DRAFT
June 4, 1993
excess margin of safety) would typically involve use of a complex simulation model
which can represent episodic loads and unsteady receiving water conditions. However,
simpler models, including one-dimensional steady state models, may be adequate in
some cases, if (1) the uncertainty introduced into the analysis by use of the simplifying
assumptions is incorporated into the margin of safety (either through an explicit
uncertainty analysis or through use of conservative assumptions), and (2) the spatial
resolution of the model is adequate to address the appropriate WQSs and impairments
of designated uses. For instance, a simple analytical mixing zone model might be
sufficient to evaluate the impact of a rapidly degrading contaminant even in a complex
estuarine hydrodynamic situation; however, a one-dimensional model could not assess
the spatial distribution of a longer-lived pollutant in such a situation. Additional
guidance on the applicability of simple models for TMDLs is provided in Chapter IV.
Also, from a practical standpoint, simple models may have to be used to establish
LAs and WLAs where resources are in short supply (e.g., there is not enough time, or
funds for sampling, or staff are short on expertise or computational power to perform
more complex modeling analyses). In these cases, however, these limiting factors should
always be made clear to the decision-maker and the potential implications of error in
decision-making based on the simpler approach should be determined. A phased
approach can always be used to follow up in such a situation should resources be freed
up in the future.
In addition to their potential use for direct TMDL development, simple loading
models can assist TMDL development by providing input for more complex receiving
water models that are in turn used to establish the WLAs and LAs. For instance, some
well recognized and frequently used receiving water models (e.g., QUAL2E, WASP) do
not explicitly include routines for simulating runoff from various land uses. In these
cases, rather than abandon these models in favor of a more complex watershed model,
it might suffice for the modeler to use one or more of the simple runoff estimate models
to arrive at discrete loads which could, in turn, be input to the existing receiving water
model.
2.4 Assessing Scoping Model Accuracy
Upon application, the model user should always assess the accu*a^ and
results. Since no model is perfect, the objective is to assess the magiu amndooint
to determine whether that degree of error is import**. From a pra^ smd^mt,
model inaccuracy is important if it causes the wrong decision »J* "a^e- * 1
models often have a large potential prediction error associated with them, dose atte
should be paid to the assessment of error on any application of significance.
2-5
-------
The Cadmus Group, Inc.
DRAFT
June 4,1993
This is typically done through a comparison with the problem setting. For
instance, the modeler may ask, "to what degree does the simple scoping model capture
the expected variability of the real world system? Are climatic conditions (i.e., wet vs.
dry year) important?," etc. If the model only produces an "average" condition prediction
and the extremes are of importance to addressing the problem, then a more sophisticated
model that captures the variability would be warranted.
Timing also may be important to model accuracy. A simple model that produces
annual average loading may suffice where receiving water impacts occur over a long
period due to high retention times and slowly reacting pollutants. On the other hand,
such a model would probably not accurately address a problem that occurs on a
seasonal or shorter term (e.g., daily, hourly) basis. This issue is covered further in
Chapter IV.
In addition, the modeler should be aware of the extent that the simplification of
a system assumed through the scoping model biases the result. For instance, a
reasonable worst-case event mean concentration (EMC) could be assumed for each
loading event based on the NURP or USGS studies for urban runoff. If the worst-case
loads are assumed to coincide with low flows in the receiving waterbody (an unlikely
proposition), such an analysis is likely to overestimate the magnitude of concentrations
seen downstream and thus biases the result. The impact of this bias is that the available
fraction of the waterbod/s LC that can be assigned for WLAs is underestimated. This
may not be an issue if there is little need for WLAs. On the other hand, it could be of
particular importance in situations where there is a large demand for WLAs and the
remaining assimilative capacity is limited or exhausted. In the latter case, a more
detailed modeling analysis could provide a more accurate estimate of NPS impacts,
thereby reducing the excess MOS that was built into the scoping model via a
conservative assumption (i.e., that worst-case concentrations occur during every runoff
event).
Where actual spatial or temporal data distributions are available for response
variables, the modeler can compare model predictions to these data for their degree of
accuracy (e.g., check for patterns of over- or underestimation, magnitude of error, etc.).
Evaluations can range from as simple as a bivariate plot comparison to more statistically
rigorous techniques such as the chi square or Kolmogorov-Smirnov two-sample tests (see
Chapter VII). The outcome of these statistical evaluations of model accuracy can answer
whether the model's level of accuracy is acceptable for decision-making or whether a
more accurate model is needed.
The following checklist may be helpful to modelers to review scoping model
accuracy:
2-6
-------
The Cadmus Group, Inc.
DRAFT
June 4,1993
• Does the model's spatial dimension match the physical characteristics of the
system of concern?
• Is the natural temporal variation of the response variable adequately covered by
the model?
• Is there any evidence of model bias? (e.g., consistent over or under prediction)
• Is the error associated with the model prediction acceptable for decision making
purposes?
In some cases, there may be a paucity of data for the modeler to use to make
these assessments. However, if that is the case, then the modeler undoubtedly had to
make gross assumptions in order to perform the modeling analysis at all. Therefore, it
is probable that the lack of knowledge led to conservative assumptions such that there
is an implicitly larger MOS. If the MOS is unacceptably large, then a more detailed
modeling analysis aimed at reducing the uncertainty and improving on accuracy would
need to be performed. See Chapter IV for more guidance on model selection,
particularly where simplified methods may be adequate, and Chapter VII for greater
detail on evaluating model accuracy.
2.5 Types of Simple Scoping Methods
A considerable amount on information regarding simplified mod®^g
is available from EPA. Specific model or method descriptions are provided in th
following references:
EPA. 1992. Compendium "f WatPrshed-Srale Models for TMDT. [
Environmental Protection Agency, Office of Wetlands, Oceans and Watershe . P°
No. EPA841-R-92-002.
Donigian, A.S., and W.C. Huber. 1991. Modeling of Noppoint gourq» Water QualitYm
Urban and Non-Urban Areas. U.S. Environmental Protection Agency. Report wo.
EPA/600/3-91/039.
EPA 1985 Water Oualitv Assessment: A Screening Procedure for TO
Conventional Pollutant^ in Surface ar^ Ground Water: Parts I & II- U.S.
Protection Agency, Environmental Research Laboratory, Athens, GA. Rep
EPA/600/6-85/002a & b.
2-7
-------
The Cadmus Group, Inc.
DRAFT
June 4,1993
Rather than repeat those specific model descriptions here, this guidance will focus
on separating these methods into general categories that can be used for different types
of applications and then discuss considerations in the use of those types of methods in
TMDL decisions.
Simple methods generally fall into two category types, analytical or statistical.
Analytical methods typically involve models that attempt a simplified representation of
the physical, chemical and/or biological components and processes occurring within a
waterbody. These types of models tend to ignore many of the real world complexities,
but they are still causally based. The Streeter-Phelps equation coupling oxygen
demanding substances to dissolved oxygen deficit in the water column is an example
of a simple analytical model. Statistical methods, on the other hand, reflect observed
relationships without explanation of causality. The regression equations developed by
USGS (Tasker and Driver, 1988) to provide storm-mean pollutant loads and
corresponding confidence intervals based upon monitoring data at 70 gaging stations in
20 States provide an example of a simplified statistical approach.
Given their relative ease of use, if s often possible to run both a statistical and
analytical scoping model and compare the results. For example, the two different
scoping methods might be used to estimate the annual nutrient loading into a lake or
estuary. The analytical approach could involve a mass balance of loads from PS and
NPS throughout title watershed, with source load estimates obtained from NPDES
monitoring data and land-use export coefficients. The statistical approach, on the other
hand, might involve the development of a regression model where flow and
concentration monitoring data are used to establish a flow/nutrient loading relationship,
and the model is used to estimate loading under an average annual flow regime. The
statistical approach may be preferred where adequate monitoring data are available,
whereas the analytical approach would be preferred in situations lacking field data.
However, it is difficult to apportion the resulting loads from the instream regression
model back to specific source categories. So if objectives extend beyond comparative
watershed loading or total loading, a more detailed method would be needed.
One advantage to the use of statistical models is that they can often be used to
calculate a frequency distribution of loading, which is particularly advantageous to
assessing the impacts of wet weather events. For example, where these models describe
a frequency distribution of event mean concentrations in loading, they can be combined
with information of the frequency distribution of rainfall to yield a frequency
distribution of loading to the receiving water. The Federal Highway Administration
(FHWA) Model is a simplified scoping model (spreadsheet format) that is built upon this
principle; summary statistics are produced on the magnitude and frequency of
occurrence of instream pollutant concentrations (more details can be found in "The
Compendium of Watershed Scale Models," U.S. EPA 1992).
2-8
-------
The Cadmus Group, Inc.
DRAFT
June 4, 1993
2.6 Examples of Simple Model Application
The following applications are offered to illustrate the points made above:
2.6.1 A "Framing" Example
Acute toxicity effects have been observed in a small watershed of a western
metropolitan area following rain events. Since the effects were noted following the
storms, no data are available for specific toxic constituents that may have been present
in the runoff during the storm. However, given the highly urbanized nature of the
watershed, officials have hypothesized that heavy metals may be the cause. In
particular, trace amounts of lead (Pb) have been found in water column and sediments
during non-storm events and, while the levels observed were not acutely toxic, they may
be residual levels and indicative of what might be washing off the land at higher
concentrations during storm periods. In order to see whether its possible that Pb could
be the cause, and before too much money is spent attempting to pursue Pb further, the
manager decides to use a simple model to estimate the an average event mean
concentration for Pb.
The USGS regression model (Driver and Tasker, 1988) developed for the western
region of the U.S. for mean storm-runoff Pb concentration was used to get a rough idea
of the probable magnitude of the stormwater Pb concentration in the watershed in
question. The equation takes the form:
Pb = 141(TRN)*,3<7(DA)'li5(LUI+l)'109(LUC+l)ao34(LUN+2)*'08<(MAR)0<6BCF
(ug/1)
(R2 = .19, Std. error * 88 %)
where: TRN = Total Storm Rainfall in inches
DA = Total contributing drainage area in square miles
LUI = Industrial land use, as % of contributing area
LUC = Commercial land use, as % of contributing area
LUN = Nonurban land use, as % of contributing area
MAR = Mean annual Rainfall in inches
BCF = Bias correction factor (1.304 for Pb in Western Region)
The storm rainfall level (i.e., TRN) of interest is determined to be approximately 0.5
inches since that is the level at which effects have been observed. The drainage area is
estimated to be about 1.0 square mile, and has a relatively high percentage of
impervious area (LA = 50 %) due to its urbanized nature. The watershed is composed
of 40 percent industrial land use (LUI), 30 percent commercial land use (LUC), and 15
percent nonurban land use (LUN). The mean annual rainfall is estimated at 10 inches.
2-9
-------
The Cadmus Group, Inc.
DRAFT
June 4, 1993
Using the regression equation, an event mean concentration (EMC) of 152 ug/1
is estimated. Comparing this to the final acute value (FAV) for Pb of 67 ug/1, the
manager sees that the model prediction far exceeds the acute toxicity criterion (even after
appropriate dilution calculations in the receiving stream). However, she's worried that
the low R-square and relatively high standard error of the model may result in too high
of a prediction error. She decides to calculate a 90 percent confidence interval and
discovers that the lower bound of 78 ug/1 is still above the acute threshold of concern.
With this information in hand, the manager decides to initiate a pilot storm monitoring
program that will include lead monitoring to further pursue the matter.
2.6.2 A Targeting and Prioritizing Example
A nutrient budget was developed for the Neuse River Basin in North Carolina for
comparison of PS and NPS loads throughout the basin. Loads were estimated for each
of 14 sub-basins vising pollutant loading factors ("export coefficients") for NPS and
NPDES discharger monitoring data for PS (NCDEHNR-DEM, 1993). Estimation of land
uses at this scale (The Neuse Basin encompasses approximately 6200 square miles) was
facilitated by the use of a geographic information system that contained data obtained
from a LANDSAT study.
The results (summarized for phosphorus in Table 2-1.) of the simple estimation
technique revealed portions of the watershed that merited further attention. For
instance, the majority of phosphorus loading in the basin can be attributed to two sub-
basins (030402 and 030407). Of these two areas, point sources are the biggest contributor
in sub-basin 030402 whereas agricultural NPS was by far the biggest estimated
contributor in 030407. The 030407 watershed is comprised entirely by the drainage to
Contentnea Creek, a major tributary to the Neuse River. The State used the nutrient
budget, along with assessment information indicating use impairment, to target
Contentnea Creek and rank it as a high priority watershed for TMDL development.
2-10
-------
The Cadmus Group, Inc. DRAFT June 8,1993
Table 2-1. Point and Nonpoint Source Phosphorus Loading to the Neuse River Basin by Land Type and Sub-basin.
Subbasin
Agriculture
(kg/year)
Forest
(kg/year)
Urban
(kg/year)
Wetland
(kg/year)
Water
(kg/year)
Otter
(kg/year)
Point Source
(kg/year)
Total*
(kg/year)
Totals
Percent
08 04 01
50,974
11,413
26,658
2,674
1957
1,341
17,603
113,620
107*
03 04 02
64,284
9,307
34,510
1,692
842
549
110,397
221,581
20.8%
03 04 03
16,173
1,299
4,999
273
97
93
8,414
31,348
2.9%
03 04 04
397»7
2,598
5,701
583
250
145
3,436
52,508
4.9%
03 04 05
68,685
6,299
1,025
371
486
568
14,884
92,317
8.7%
03 0406
43,070
3,962
3,369
344
174
198
1,919
53,036
5.0%
03 0407
145,823
U699
1,269
699
661
650
24,614
186,414
17.5*
03 04 08
19/165
3,062
1,091
883
481
541
157
25,680
2.4*
03 04 09
30,570
5,267
186
771
45
530
415
37784
3.5%
09 04 10
18,311
6,971
3,328
2,703
31,875
1,506
24772
89,467
8.4%
03 0411
27,569
4,920
2.457
1727
152
1,926
405
39,156
3.7%
03 0412
19,459
2,036
400
163
536
108
16,026
38728
3.6%
03 0413
6,377
732
1,564
1,517
24,541
357
0
35,088
3.3%
03 0414
1,554
268
296
1,207
45,570
148
0
49,043
4.6%
Totals - kg/yr
552,111
70,832
86354
15,608
108,666
8,659
223,043
1,065772
Totals-%
51.8*
6.6*
8.1*
1.5*
10.2*
as*
20.9*
100.0%
The following subtotals are for the Neuse Basin below Fills Lake Dam to New Bern (all subbasins except 01,13 and 14)
2-11
-------
The Cadmus Group> Inc.
DRAFT
June 4, 1993
References. Chapter 2
Donigian, A.S. Jr. and W.C. Huber. 1991. Modeling of Nonpoint Source Water Quality
in Urban and Non-Urban Areas. EPA/600/3-91/039. ERL, Athens GA.
NCDEHNR-DEM. 1993. Neuse River Basinwide Water Quality Management Plan.
Water Quality Section.
Tasker, G.D. and N.E. Driver. 1988. Nationwide Regression Models for Predicting
Urban Runoff Water Quality at Unmoiiitored Sites. Water Resources Bulletin 24(5): 1091
-1101.
U.S. EPA. 1992. Compendium of Watershed-Scale Models for TMDL Development.
EPA841-R-92-002. Office of Water.
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
Chapter III. Simulation Models: Model Definitions and
Available Models
As discussed in Chapter I, models are used to help establish the links between
pollutant loads and WQSs in a receiving waterbody, with the basic objective of obtaining
compliance with the CWA. Where any nonpoint sources (and particularly episodic, wet-
weather sources) are involved in the impairment of a waterbody, the effects of nonpoint
BMPs or point source load reductions on WQS excursions is difficult to predict
quantitatively from past observations. The TMDL developer will often need to employ
modeling to extrapolate beyond the monitoring record and to assess the potential
impacts of proposed control strategies.
The main purpose of this chapter is to discuss the types of models that may be
used in the estimation of TMDLs, and to provide a formal basis for their evaluation and
comparison in terms of the needs of a TMDL developer at a specific site. It thus focuses
on model type and availability, but also provides a catalog and reference to a number
of useful models supported (to some extent) by Federal agencies. In Chapter IV we will
come at the problem from the other direction, that is, given the characteristics of the
TMDL problem and site, what are the characteristics of appropriate models. Chapter IV
represents the process of model identification. Comparing the desired set of
characteristics to the set of available models (this chapter) constitutes model selection.
While this chapter provides general information on the characteristics of
simulation models, and also lists specific models available for various purposes, the
focus is on model characterization. We do not provide comprehensive reviews of
individual models. This is for two reasons. First, detailed reviews of most of the
appropriate models are available in other EPA guidance (notably including, for loading
models, "Compendium of Watershed-scale Models for TMDL Development" (U.S. EPA,
1992) and "Modeling of Nonpoint Source Water Quality in Urban and Non-urban Areas"
(Donigian and Huber, 1991), and, for receiving water models, the WLA guidance).
Secondly, the status and availability of simulation models is subject to continual change,
and models other than those listed here will also likely be appropriate for TMDLs.
3.1 Types of Models Used in TMDL Development
Simulation models can be used to assess aspects of NPS loading, pollutant fate
and transport in receiving waterbodies, and ecological and biological impacts within the
receiving water. The exact goals of modeling will be deterpiined by the need to evaluate
the sources and control of waterbody impairment, expressed in terms of the applicable
3-1
-------
The Cadmus Group> Inc.
DRAFT
June 8, 1993
numeric or narrative criteria. Specific types of impacts most likely to be encountered in
a given waterbody type are discussed in Section 5.6.
Thus, estimation of TMDLs can require modeling or analysis of a number of
different physical domains or processes. (In the most general sense, we will use
modeling to refer to anv sort of structured analysis of a domain or process; sorting out
whether a complex simulation model or a simple scoping or empirical analysis is to be
used is a subsidiary question). For instance, to investigate the effects of agricultural
nutrient loading delivered via a stream into a lake we might need to consider the
following components:
1. Buildup and availability of nutrients under current agricultural practices,
2. Patterns of overland flow during storm events;
3. Transport of nutrients in washoff during overland flow events;
4. Flow regime in the receiving stream;
5. Transport of nutrients in the receiving stream;
6. Delivery and mixing of the nutrient load into the lake;
7. Water circulation and nutrient cycling within the lake;
8. Algal population/nutrient responses to the nutrient loading.
In most cases, not all of these aspects will (or can) be modeled in equal detail, and
some may be bypassed with "appropriate assumptions". Various of the individual
aspects may be addressed together within a single simulation model (or simple screening
approach). However, in most cases there will not be any single, "as is" modeling
package available to meet all, the needs of a TMDL, particularly as the different
components may need to be addressed at different levels of detail. To continue the
nutrient loading example, the TMDL developer (at a fairly simple level of analysis)
might need to combine (1) an empirical or loading function analysis of monthly nutrient
delivery from agricultural land use, (2) a simple, mass-balance/dilution calculation of
transport within the stream, (3) a seasonal model of nutrient cycling within the lake, and
(4) an empirical model of lake nutrient response. This example combines four rather
simple models. On the other hand, if the concern was with acute toxic effects of
pesticides, it might be necessary to estimate or model (1) pesticide availability, (2)
washoff during individual storm events, (3) dynamic transport and reactions within the
stream, and (4) mixing and decay near the outfall into the lake. All four components
could involve rather complex models, if sufficient resources were available to undertake
such an analysis.
3.1.1 Taxonomy of Simulation Models for TMDLs by Physical Domain
Because a modeling strategy for estimating TMDLs usually will involve more that
a single modeling component, and because TMDLs are applicable to all types of
3-2
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
receiving waterbodies, this guidance must address a wide range of simulation models.
For the purpose of discussion it is convenient to start with a formal categorization or
"taxonomy" of the types of models that may be useful for TMDLs. Models may be
classified on many different criteria, including types of pollutants considered, level of
complexity, etc. However, the most natural division is on the basis of the physical
domain addressed. The basic categorization by physical domain of types of models
useful for TMDLs involving nonpoint sources is shown in Box 3-1.
It is important to emphasize in this classification that water flow ("quantity") and
pollutant transport ("quality") constitute separate aspects of both nonpoint source and
receiving water simulation. Indeed, these aspects are often addressed by different
models or approaches. A common situation is that in which a hydrodynamic model is
first run to describe receiving water flows. The results are then used to drive a separate
water quality model. The analyses may be at rather different levels, e.g., we might
require only a steady state description of the hydrology (perhaps a design flow) but then
superimpose a time-dependent model of episodic loads of a reactive pollutant For
another example, we might consider analysis of a storm sewer system, in which a
detailed hydrodynamic simulation is made of the flows, and pollutant transport is then
estimated as an average concentration in flow based on drainage area land use - thus
combining a complex model of flow with a very simple "model" of pollutant transport.
3-3
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
Box 3-1. TMDL Model Taxonomy by Physical Domain
1. Loading Models. Used to estimate urban and non-urban nonpoint source-
derived loading to receiving waters and to predict BMP effectiveness. These
models can generally address episodic, rainfall-dependent loads; some also address
dry weather flows. There are two major subclasses:
A. Urban Loading Models, which include combined sewer overflow and
urban stormwater discharge models. These are generally point
sources, in terms of discharge to the receiving waterbody through
outfalls, which, however, include consideration of nonpoint washoff
into the collecting system.
B. Non-urban Loading Models, which address runoff from agriculture,
forestry, suburban development, and other distributed runoff sources.
For either urban or non-urban loading, models can address some or all of
the following aspects:
i. Runoff hydrology: the generation and routing of overland, sewer,
and drainage flows.
ii. Source availability: the buildup and availability for washoff of
pollutants and sediment.
iii. Sediment transport: The movement and delivery of sediment (and
sediment bound pollutants) from sources to the receiving waterbody.
iv. Chemical transport: the movement, reactions and delivery of
chemical pollutants (dissolved or sorbed) to the receiving waterbody.
2. Receiving Water Models. Used to estimate receiving water flow and
transport, and responses to pollutant loading, and to establish TMDLs/WLAs/LAs
to meet water quality standards. Three general types of physical domains are
addressed, distinguished by the differing importance of advective and dispersive
transport processes. An additional distinction may be made between far-field and
near-field (local mixing zone) models:
Far-field Models
A. Rivers and streams, in which unidirectional advection is the
dominant transport process.
3-4
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Box 3.1 (continued)
B.
Lakes and reservoirs, in which flow is not unidirectional and
macrodispersion is often the dominant transport process.
C.
Estuaries, in which both advection and dispersion (driven by tidal
fluxes) are important.
Near-field Models
D. Mixing Zone Models, which consider near-source dilution only.
Within each of these physical model types, receiving water models may
address some or all of the following:
i.
Hydrology, or the movement of water.
4 *
11.
Sediment transport, scour and deposition.
• • ¦
111.
Chemical transport and reactions.
3. Ecological Receptor Models. Used to predict ecological responses to
pollutant loading to a receiving waterbody. Their "physical" domain thus may be
the ecosystem. These may be needed to establish TMDLs/WLAs/LAs to meet
certain narrative criteria for the protection of aquatic life, habitat criteria and
biocriteria. Two classes of Ecological Receptor models are commonly used in
TMDLs:
A.
Habitat Modification Models, which may involve consideration of
stream flow dynamics or stream morphology (e.g., channel down-
cutting, bank erosion, bar deposition, etc.) in the context of habitat
suitability.
B.
i.
ii.
Biological Community Models, under which we may include:
Population dynamics (or biomass production) models, which might
address phytoplankton, zooplankton, periphyton, benthic
macroinvertebrates, fish, or piscivores or other animals dependent on
the aquatic food chain.
Biodiversity or ecological indicator models.
3-5
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
3.1.2 Other Bases for Model Classification
Physical domain is perhaps the most obvious means of classifying models, but is
not the only basis for classification. That is, if we are modeling impacts in a lake, we
don't simply choose among all possible lake models at random. Instead, other issues
must be considered (e.g., form of relevant WQSs, cost of model implementation, etc.),
and all these other considerations can also be used to classify available models. As
noted above, Chapter IV will consider the issue of model identification, while this
chapter is concerned with setting the stage for model selection, in which the user must
evaluate which among many potential models and modeling approaches provide a good
match to the modeling requirements of the site. The other bases for model classification
which are relevant to the issue of model identification will also need to be considered.
Among these are, in addition to physical domain:
• Temporal Representation and Scale. Models may be classified as steady
state or dynamic in their representation of a given process. Steady state
models do not include a derivative with respect to time in their
formulation; that is, they represent the ultimate response to a steady
forcing function given infinite time. Dynamic models represent temporal
variability. Some simulation packages may combine steady representations
of some components with dynamic representation of others - for instance,
steady flows with time varying loads superimposed. Where
dynamicprocesses are represented there are usually limits on the timesteps
or temporal representations that are appropriate for a given model.
Numerical models often have a maximum time step determined by
conditions of solution stability. On the other hand, lumped parameter
models usually have a minimum time step below which the simplifying
assumptions used in model development do not permit accurate
representation.
• Spatial Representation and Scale. The physical representation employed
by a simulation model may be in zero, one, two or three dimensions (zero
dimension refers to receiving water models which treat the receiving
waterbody as a completely mixed tank or continuously stirred reactor).
Similar to the temporal representation, there are usually limits on the
spatial increments that are appropriate for a given model. Obviously, a
model employed in the estimation of a TMDL should be able to produce
results at a spatial scale appropriate to the WQSs under consideration. For
instance, in lake eutrophication problems, die spatial dimension of interest
may be the whole lake (or whole epilimnion), while, on the other hand,
consideration of CSO discharges of fecal doliforms may require a fine
spatial scale to assess impacts on beaches, drinking water intakes, etc.
3-6
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
• Constituents and Processes Simulated. Simulation models have differing
abilities to represent types of constituents. Distinctions may include
conservative vs. degrading, dissolved vs. sediment sorbed, organic vs.
inorganic. Some models only handle one group of constituents (e.g., DO
and BOD). Besides chemical contaminants, TMDLs may also need to
consider modeling of sediment itself, temperature, and habitat and
geomorphological parameters.
• Practical Considerations and Costs. Besides their ability to represent
physical processes, models differ widely in their ease of use, data
requirements, and the level of effort required for successful
implementation. These are all valid considerations in the selection of a
simulation model for use in TMDLs.
The various means of classifying models will be revisited in Section 3.3, which
provides a summary (and classification) of various federally supported models useful
for the calculation of TMDLs. This provides a language, or formal set of criteria for
describing which models are useful in which situations. In Chapter IV, the identification
of appropriate models will be developed using the same set of criteria.
3.2 Introduction to Simulation Models
This section provides a brief introduction to the major concepts and terminology
of simulation modeling relevant to TMDLs. This is essentially an annotated glossary of
concepts useful for the discussions which follow.
3.2.1 Typical Forms of the Governing Equations for Flow
Simulation models which attempt to provide a causal, physically-based
representation of flow and transport (as opposed to empirical or statistical formulations),
are generally based on the principle of conservation of mass, and, where appropriate,
the principle of conservation of energy or momentum, expressed through partial
differential equations. Performance of a simulation model is dependent on the way in
which these equations are formulated, and the way in which the equations are solved.
Equations of Flow. All physically based descriptions of flow consider
conservation of mass. That is, for a control volume of a simulation, input of water
minus output must equal storage within the control volume. Conservation of mass
alone is sufficient to describe certain processes, such as reservoir storage under gradually
changing inflows. However, flowing water also possesses energy or momentum, and
the principal of conservation of energy must also be considered to develop a complete
3-7
-------
The Cadmus Group, Inc. DRAFT June 8, 1993
description of flow, particularly where gradients are relatively large or changing over
time.
For flow in a channel, the complete equations of flow are typically represented
by the St. Venant equations. The St. Venant equations may apply in three dimensions,
but can be written in one dimension as
— + V— * S. + 2^! = a - a, ~ a
St & A St A • ' *
i2 ~ M , 9l
& at *'
where
V is the velocity of flow
t is time
x is distance along the axis of flow
g is the gravitational constant
A is cross sectional area
y is the depth of the centroid of flow
q, represents net flows into the channel for a given control volume
ag is acceleration due to gravity
af is acceleration due to friction
aw is acceleration due to wind
Q is the flow through the control cross section (= V*A)
The first equation is the conservation of momentum or energy equation, with 7
terms, while the second equation is the continuity or conservation of mass equation, with
3 terms, for a total of 10 terms. In order, these 10 terms represent for a given control
section:
1 and 2. Taken together, these terms represent the total differential of velocity
with respect to time, i.e., the local acceleration.
3. represents the hydrostatic pressure change across the section.
4. represents the component of momentum associated with inflows to the section.
5. is the acceleration due to gravity.
6. is the acceleration due to bed friction.
3-8
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
7. is the acceleration due to wind.
8. is the change in flow across the section.
9. is the change in storage in the section.
10. is the lateral inflow volume into the section.
A solution of a particular form of these equations yields a method of flow routing.
In many situations, many of the terms within the momentum equation are relatively
insignificant on an order of magnitude basis. For steady state analysis, the partial
derivatives with respect to time are assumed to be zero. The representations of channel
flow may be classified on the basis of which terms are included or omitted.
Hydrologic Routing methods make use of the continuity equation only; that is, they
implicitly assume that changes in energy need not be considered to model the changes
in flow. This is usually adequate to describe longer term averages, and situations in
which gradients do not change rapidly. Hydrologic routing is contrasted to hydraulic
routing, in which some form of the momentum equation is included. Typical methods
of hydrologic routing include the Muskingum method and the SCS Convex Method for
channels and the Storage Indication method for reservoir routing.
Kinematic Wave Routing considers the momentum equation, and is thus a hydraulic
routing method. However, only the most significant terms of the momentum equation
are introduced and in steady state: it is assumed that acceleration due to gravity and due
to body friction are in balance, that all other terms of the momentum equation are
insignificant, and that the only time derivative present is that in the continuity equation.
The method is thus applicable to uniform water surface profiles and (locally) steady
flows, although discontinuities in the flow regime are considered. Kinematic wave
methods can estimate the rate of propagation of flood waves or changes in flow and are
more sophisticated than hydrologic routing methods. However, they are more
commonly used for overland flow than channel flow. The SCS (1979) Att-Kin TR-20
method represents a combination of storage indication and kinematic wave routing
techniques for channel flow.
Dynamic Wave Routing methods address unsteady and nonuniform flow by
including acceleration and internal momentum energy terms. This is necessary to obtain
accurate representation of flow effects in which the friction slope of the water is not
equal to the bed slope, such as occurs in backwater effects. Various representations are
possible, and not all the terms of the momentum equation are always included. For
channel flow, the effects of hydrostatic pressure change (overpressure) across the control
volume (term 3 above) are often ignored, as in the formulation for the hydrodynamic
3-9
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
model DYNHYD in the WASP package (Ambrose et al., 1988). Complete dynamic wave
formulations are important for the study of extreme flood events, and are incorporated
into several National Weather Service models (e.g., DWOPER).
The above terminology is most commonly applied to streams, rivers and estuaries.
The same physical principles apply to overland flow, and kinematic wave methods are
often used to describe the time history of flow generation on small, and particularly
impervious areas. However, a detailed description of the energy component of overland
flow from a larger, heterogeneous area is usually neither practical nor relevant For
TMDL development we will usually be interested in total runoff generated from a land
area during a precipitation event, or, at most, an approximate time history of runoff.
Therefore, overland flow is usually addressed via empirical or mass balance methods.
The most commonly used empirical approach is the SCS Curve Number Procedure (SCS,
1968). This predicts runoff from a storm based on antecedent conditions, soil
classification, and type of cover. Mass balance approaches generally estimate runoff
based on precipitation minus infiltration, and employ a model of the time-dependent
decay (and recovery) of infiltration capacity.
Simulation of flows in lakes often involve some rather different problems. This
is because lakes can present very different dimensions from flowing streams, and
typically show three dimensional patterns of flow, often with thermal stratification. In
most lakes, wind is the principal mixing mechanism, and temperature gradients the
principal resistance to mixing. Detailed simulation of the internal flows of lakes requires
consideration of the thermal energy balance and density gradients due to temperature
and dissolved solids.
3.2.2 Equations for Sediment and Pollutant Transport
Simulation of sediment and pollutant transport is usually built atop a description
of flow, which may either be derived from a flow model, or described from observation.
Transport processes include advection, which is the movement of constituents with the
bulk flow of water, and dispersion, which is the result of mixing processes. In practice,
dispersion may be thought of as containing the net results of all those flow processes
which occur at a scale too small to be captured by the flow model, as well as mixing due
to molecular diffusion. Transport is expressed via an advection-dispersion equation.
In one dimension (i.e, averaged over the cross sectional area of a channel) the
advection-dispersion equation can be written in the following form (using the notation
of Ambrose et al., 1988):
3-10
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
S(AO m _ WAQ + ^
|£] * A (SL+SJ *A S,
dt cbc cbd
where
A is the cross-sectional area (L2)
C is concentration of the water quality constituent (M/L3)
t is time (T)
x is distance in the direction of flow (L)
Ux is the longitudinal advective velocity (L/T)
^ is the longitudinal dispersion coefficient (L2/T)
Sl is the direct and diffuse loading rate (M/L3-T)
Sg is the boundary loading rate (including upstream, downstream, benthic
and atmospheric loading; M/L3-T)
Sk is a net source or sink rate, representing reactions and transformations
A similar representation may be made in two or three dimensions as required.
A separate equation is required for each constituent Often, these are linked through the
last term, representing transformations including internal sources and sinks. For
instance, the equations for transport of BOD and DO are Jinked through a term in which
the decay of BOD causes a corresponding removal of dissolved oxygen (or increase in
oxygen deficit).
Depending on the process, a variety of kinetic formulations have been proposed
in the literature. However, in most surface water quality assessments, chemical
transformations are approximated by a first order rate law with rate constant k of the
form
which is equivalent to an exponential decay process.
Another important constituent rate determination is for the boundary loading
rates (distinct from advective and dispersive transport). For instance, in simulating
sediment these include exchange across the benthic boundary representing losses due
to deposition and gains due to scour. For many dissolved constituents we may need to
consider volatilization into the atmosphere and sorption on to suspended sediment.
EPA provides references on typical dispersion coefficients and rates and constants
for conventional pollutants in Bowie et al. (1985) and for conventional and
nonconventional pollutants in Mills et al. (1985). Another important reference for
estimation of chemical reaction and transformation rates is Lyman et al. (1990).
(M/L3-T).
3-11
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
3.2.3 State Variables, Boundary Conditions, and Initial Conditions
State Variables are those variables whose behavior a simulation model attempts to
reproduce and predict. Choice of state variables is a critical part of model
implementation. The more state variables that are included, the more difficult the model
will be to implement and calibrate as the model is likely to be over-specified relative to
the data; however, if important state vari )les are omitted from the simulation the model
may produce unrealistic results, or be unable to answer necessary questions for the
TMDL. For instance, if DO is being simulated, should phytoplankton be included as a
state variable (i.e., should phytoplankton biomass be simulated), or should it be
represented as a constant DO source/sink? The answer depends in part on whether
proposed load allocations to control oxygen demand are also likely to affect
phytoplankton density, and thus affect DO. The general rule proposed by Thomann (in
Freedman et al., 1992) is: "Keep state variables to a minimum; model only those for
which data exists; but always include those state variables which will be impacted by
a WLA." However, while it is desirable to model only state variables for which data are
available, this is not always appropriate, as Thomann himself notes, when unmonitored
components are essential to the understanding of key state variables. For instance, both
dissolved and particulate phases must generally be included.
Boundary Conditions and Initial Conditions constrain the partial differential equations
from a general to a particular solution. Dynamic flow and water quality problems can
be thought of as initial boundary value problems, in which a solution is propagated in
time from an initial condition but also depends continuously on boundary conditions.
In contrast, steady state solutions may be boundary value problems, which do not
depend on initial conditions.
Boundary conditions, as the name implies, describe the state variable along the
boundaries of the solution domain at all times during the simulation. The boundary
conditions may constitute the specification of prescribed values at the boundary (a first-
type or Dirichlet boundary condition), the specification of the normal derivative or
gradient on the boundary (second-type or Neumann boundary condition), or the mixed
specification of a function of the state variable and its derivative on the boundary (third-
type or Robin boundary condition). For instance, in simulating hydrology of a coastal
plain river in one dimension in which channel cross-sectional area is a state variable, we
might specify a time series of inflows at the upstream end of the simulation grid and
tidal elevation at the downstream end. This specifies a rate of change of cross-sectional
area upstream and a specified cross-sectional area at the downstream boundary. Initial
conditions are also required for solution of time propagation models. In some cases
(e.g., river flow simulations) the solution after a few days simulation will be rather
insensitive to initial conditions. However, in other cases, such as water quality
3-12
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
simulation for a slug discharge into a lake, the predicted time history may be highly
sensitive to initial conditions of circulation and background concentrations.
3.2.4 Solution Methods for Numerical Models
A simulation model is specified by writing the partial differential equations and
specifying an appropriate set of initial and boundary conditions. These must be
combined into a solution form. The behavior of a simulation model is determined not
just by its specification but also by the method of solution of the governing equations.
In terms of choosing an appropriate model, the solution method is of interest primarily
because of restrictions it may place on the representation of spatial and temporal
variability or, conversely, the maximum time or spatial increment that can be used in
simulation.
For very simple model specifications (generally those where the parameters and
fluxes are constant in space and time) it may be possible to derive an analytical solution,
which allows a direct and exact analysis of the model. A good example is provided by
the classic Streeter-Phelps equation, which predicts dissolved oxygen deficit downstream
of a constant source in terms of reaeration, BOD deoxygenation and BOD loss rates.
Analytical solutions are often useful for screening, due to their ease of implementation.
Because the solution is exact, they are also not subject to instability from numerical
roundoff errors. However, the usefulness of the simple formulation is limited by its
assumptions of constant loading, constant flow, and constant reaction rates.
To solve a more complex model specification, i.e., one where parameters vary in
space and/or time, it is necessary to resort to numerical solution methods. Most of the
more complex quantity and quality models encountered in TMDL development use an
explicit finite difference method. In this method, the partial differential equation is solved
directly. Each partial differential is represented by a difference approximation. This is
equivalent to representing the continuous problem by a step-function or gridded
approximation. For instance, we might form a difference approximation for a partial
differential with respect to time evaluated at time t as the "slope" of the line between
discrete Values at time t and time t+1:
cbc m ~ xt
dt bt
In this case we have used a forward difference operator (over time step At); if we
evaluated from time t-1 to t+1, we would have a central difference operator (i.e.,
centered at time t), and if we evaluated from t-1 to t we would have a backward
difference operator. Of course, the approximation introduces a truncation error into the
3-13
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
solution (see Figure 3-1), and the magnitude of this error can be evaluated through
rearrangement of the Taylor series expansion (see, for instance, Anderson et al., 1984).
Once the partial differentials are expressed in an appropriate finite difference
form, we have an algebraic expression in terms of current, future (and possibly past)
values of the parameter of interest. If the algebraic expression is rearranged to solve for
the future value in terms of the known current and past values, this yields an explicit
finite difference solution, e.g.
c - f C - 3*2 ~ c At
When an explicit finite difference scheme is used, there are constraints imposed
on solution stability. In general, a solution is unstable if the approximation errors tend
to magnify as the solution is propagated; it is stable if approximation errors tend to
damp out. Methods are well developed for the analysis of finite difference solution
stability. For instance, in a first order wave equation in one dimension solved with a
central difference representation of the time derivative, the stability criterion is given by
the classic Courant-Friedrichs-Lewy condition, which requires that the absolute value of
the Courant number, cAi/Ax, be less than one, where c is the wave speed or celerity.
The stability criterion thus typically places limits on the ratio of the time step to spatial
step - i.e., to achieve a certain degree of resolution in the spatial grid, the user must
employ a time step that is below a certain critical value. The documentation for models
employing explicit finite difference solution methods should provide an analysis of the
computational stability criteria*
An alternative to explicit finite difference solutions is the implicit finite difference
solution method. In this approach, the values at all locations on the spatial grid for the
next time step are solved simultaneously as a system of algebraic equations. Implicit
finite difference equations are generally unconditionally stable, and truncation errors do
not magnify without bound, thus allowing use of a larger time step. However,
truncation errors are still present and degrade the quality of the solution as the time step
grows larger. In the implicit method, this is usually expressed as dissipation or artificial
dispersion, in which the numerical method results in the "smearing" of a wave front that
is properly sharply defined. Thus an implicit solution does not remove the burden of
finding an appropriate simulation timestep. Implicit solution methods have been less
commonly used in water quality simulation problems because of the high computational
burden of solving equations for all points simultaneously.
An alternative numerical method to the finite difference approach is the finite
element method. Finite elements work with an integral formulation of the problem,
rather than directly with the partial differential equations. Typically, a weighing
3-14
-------
Figure 3-1. Finite Difference
Approximation to a Curve
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
function is derived based on the condition that the weighted average of the residuals of
the approximate solution at the grid nodes will be zero. This method has been
extensively developed in mechanical engineering, and has received significant
application to groundwater flow problems (see Istok, 1989). Application to surface water
problems has been limited; however, finite element solutions have been used for the
sediment-contaminant transport model SERATRA (Onishi and Wise, 1982).
3.3 Characteristics of Available Simulation Models
As noted in the introduction to the chapter, this guidance is not intended to
provide detailed reviews of all available simulation models useful for estimating TMDLs.
Instead, the focus is on the development of an appropriate modeling strategy, which
identifies the necessary characteristics for whatever models are employed (see
Chapter IV). Therefore, available simulation models are presented in terms of salient
characteristics which can be matched to the requirements determined in the model
identification process. The TMDL developer should treat this section as primarily an
example of how to logically classify available models, rather than an exhaustive list of
the universe of models.
That said, this section does attempt to provide a fairly comprehensive list of a
certain subset of models. This subset is essentially those models which are (1) non-
proprietary, (2) readily available at no or minimal cost, and (3) adequately documented
and subject to some degree of support. These criteria mean that we consider primarily
simulation models developed and supported by Federal agencies, with the addition of
a few models distributed by universities and State agencies. We have purposefully
omitted commercially available models, which in many cases represent state-of-the-art
in terms of ease of use and presentation of results, but typically represent user-friendly
enhancements of public-domain models. While such models may require a significant
initial cost, they should not be ignored by the TMDL developer, where appropriate. (For
example, a number of sophisticated commercial products are available for the simulation
of storm water and combined sewer flow). Also omitted from the lists are models which
do not seem to be in current use or whose availability appears to be in question. Finally,
the same sort of criteria are relevant to the development of site-specific simulation
models.
The presentation of available models is organized according to the major subheads
of the model taxonomy displayed in Box 3-1. We discuss loading models and receiving
water models, and, under each of these heads, present the characteristics related to (1)
the flow of water or hydrology, and (2) the transport of sediments and pollutants. We
have not provided a survey of ecological receptor and habitat models, because of the
diverse character of such models, which does not lend itself to a presentation of general
3-16
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
applicability. However, the use of such models is discussed further iri Chapter VI. This
results in for tables of model characteristics, covering Runoff Quantity, Runoff Quality,
Receiving Water Quantity, and Receiving Water Quality. Of course, many simulation
models cover more than one category of simulation - for instance, combining simulation
of rural NPS flow and loading.
3.3.1 Models of Point and Nonpoint Wet-weather Runoff
Table 3-1 surveys models for Runoff Quantity (including models for rural NPS
and urban point sources, such as CSOs and stormwater, which result from the collection
and point discharge of episodic, wet-weather flows.) The table provides the following
information:
Col. 1 lists the model name.
Col. 2 lists the other general categories of simulation addressed by the software
package, and is repeated throughout the following tables. NR stands for wet-weather
runoff (both nonpoint and point) quantity or flow simulations (this table); NQ stands for
wet-weather runoff quality simulation; RF stands for receiving water flow simulation;
and RQ stands for receiving water quality simulations.
Col. 3 provides a key to the main reference for the most recent release of a given
model (keyed to the bottom of the table).
Col. 4 provides a key to reviews of a given model in four selected sources, three
of them EPA guidance.
Col. 5 indexes the land use applications of a given model, with U standing for
urban and R for rural.
Col. 6 indicates the method of simulation of flows, whether overland or collection
system. The major categories are (a) empirical runoff coefficient methods, (2) SCS curve
number ("Curve Number") methods, (3) water balance methods, based on the principle
of conservation of mass, without hydraulic simulation of momentum; (4) kinematic wave
methods, which include a simplified representation of the energy equations; and (5)
dynamic wave methods, which address (more or less) the full momentum equations.
Col. 7 indicates the temporal resolution, or time step which can be achieved by
a given model. There are two issues here: first, whether a model is applicable to
continuous simulation of flows or just response to individual events, and second the
minimum time step representation which can be reasonable achieved.
3-17
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Col. 8 indicates the agency supporting a model. Most EPA models have been
supported by the Center for Exposure Assessment Modeling (CEAM) at the
Environmental Research Laboratory, Athens. Other agencies cited include the Federal
Highway Administration (FHWA); U.S. Geological Survey (USGS); Army Corps of
Engineers Hydraulic Engineering Center (HEC) in Davis, CA; Army Corps of Engineers
Waterways Experiment Station in Vicksburg, MS (WES); the Agricultural Research
Service of the USD A (USDA/ARS); and the Soil Conservation Service (SCS); as well as
several state agencies.
3-18
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 3-1. Runoff Quantity Models
Model
Other
Uses
Main
Ref.
Reviews
Land Use
Simulation
Type
lime Step
Agency
EPA Statistical
NQ
U
a,b/C
U,R
Runoff Coeff.
Annual,
Event Ave.
EPA
USGS Regression
NQ
3
a,b
U,R
Regression
Annual,
Event Ave.
USGS
FHWA
NQ, RQ
2
a,b
Highway
Runoff Coeff.
Annual,
Event Ave.
FHWA
GWLF
NQ
4
a
U,R
Curve Number
Continuous
Monthly
Cornell Univ.
AGNPS
NQ,RF,
RQ
5
a,b
R(Ag)
Curve Number
Continuous
Hourly
USDA/
ARS
STORM-RWQM
NQ,RF,
RQ
6
afrfiA
U
Runoff Coeff./
Curve #
Continuous
Hourly
HEC
ANSWERS
NQ
7
a,b
R
Water Balance
Event
Univ.
DR3M
NQ
8
a^
U
Kinematic
Wave
Continuous
Subhourly
USGS
SWRRBWQ
NQ,RF,
RQ
9
a,b
R
Curve #/ Water
Balance
Continuous
Daily
USDA/
ARS
SWMM
NQ
10,11
aA>^4
U
Kinematic &
Dynamic Wave
Continuous
Subhourly
EPA} CEAM
3-19
-------
The Cadmus Group, Inc. DRAFT June 8,1993
¦
Table 3-1 (continued)
HSPF
NQ,RF,
RQ
12
a,b,c,d
U,R
Water Balance,
Hydrologic
Routing
Continuous
Subhourly
EPA/ CEAM
Auto Q-ILLUDAS
NQ
13
a
U
Water Balance
Continuous
Event
Illinois State
Water Survey
CREAMS
NQ
14
a4
R (field
scale)
Water Balance
Continuous
Daily
USDA/ARS
TR-20
RF
15
R
Curve Number
Event, Sub-
event
SCS
HEC-1
RF
16
R
Multiple (UH.to
Kinematic)
Event, Sub-
Evpnt
HEC
TR-55
17
U
Curve Number
Event
SCS
Key to References
1- Hydroscienee, 1979
2. Driscoll el al., 1990
3. Driver and Tasker, 1988
4. Haith el al., 1992
5. Young et al. 1986
6. HEC, 1977a
7. Beasley & Huggins, 1981
8. Alley and Smith, 1982a
9. Arnold et al., 1991
10. Huber & Dickinson, 1988
11. Roesner et al., 1988
12. Johanson et al., 1984
13. Terstriep et al., 1990
14. Knisel, 1980
15. SCS, 1973
16. HEC, 1985
17. SCS, 1986
Key to Reviews
a. U.S. EPA, 1992
b. Donigian & Huber, 1991
c. WPCF, 1989
d. McKeon & Segna, 1987
3-20
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
3.3.2 Models of Point and Nonpoint Wet-weather Loading
Table 3-2 discusses Runoff Quality Models, i.e. the transport and loading of
pollutants from episodic, wet-weather sources. The first five columns are identical to
those in Table 3-1. The additional columns summarize the following characteristics:
Col. 6 summarizes the types of pollutants which can be addressed by a given
simulation model. Some models are rather general in their potential application. Others
are characterized as applicable to nutrients (N), oxygen and oxygen demand (O), metals
(M), conservative organic pollutants (C), and nonconservative, reactive organic pollutants
(NC).
Col. 7 addresses the manner in which pollutant loads are simulated. Most models
use one of two methods: Loading Functions, in which event mean concentrations are
empirically related to land use; and buildup-washoff formulations, in which the time-
dependent availability of pollutants is attempted to be simulated. Several models rely
on sediment potency factors; i.e. pollutant load is based on a fixed fraction of sediment
scoured. This method is generally available in any model which simulates sediment
erosion.
Col. 8 summarizes sediment transport in overland flow. Many models use the
Universal Soil Loss Equation (USLE) or the Modified USLE (MUSLE). Other models
attempt to simulate sediment detachment and transport by physical processes.
Col. 9 indicates the time step achievable by the runoff quality routine, analogous
to the time step presented in Table 3-1.
Col. 10 indicates whether the model is capable of addressing pollutant routing
(timing within the runoff)/ transformation and degradation during transport.
Col. 11 states the supporting Agency, as in Table 3-1.
3-21
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 3-2. Runoff Quality Models
Model
Other
Uses
Main
Ref.
Review
Land
Use
Consti-
tuents
Load
Generation
Sediment
Erosion
Time Step
Routing -
Transfor-
mation
Agency
1 EPA Statistical
NR
U
a>,c
U,R
General
Loading
Function
USLE/
MUSLE
Annual,
Event Ave.
no
EPA
USGS Regression
NR
3
a,b
U
N,0,M,C
Loading -
Regression
N/A
Annual,
Event Ave.
no
USGS
FHWA
NR, RQ
2
a,b
High-
way
N,C,M
Loading -
Median
Cone.
N/A
Annual,
Event Ave.
no
FHWA
Watershed
8
a
U,R
General
Loading
Function
USLE
Annual
no
USGS
GWLF
NR
4
a
U,R
HS
Loading
Function
MUSLE
Continuous
Monthly
no
Univ.
AGNPS
NR
5
ajb
R(Ag)
N,S
Potency
Factors
MUSLE
Continuous
Hourly
no
USD A/
ARS
STORM-RWQM
NR,RF,
RQ
6
diJo/iA
U
N,0,M,S
Buildup-
Washoff
USLE
Continuous
Hourly
no
HEC
ANSWERS
NR
7,11
a,b
R(Ag)
N3
Potency
Factors
Detach-
ment
Event
yes
Univ.
DR3M-QUAL
NR
13
aJbjc
U
NACM
Buildup-
Washoff
MUSLE
Continuous
Subhourly
no
USGS
SWRRBWQ
NR,RF,
RQ
9
aJb
R
S,N,C,NC
Buildup-
Washoff
MUSLE
Continuous
Daily
yes
USDA/
ARS
3
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 3-2 (continued)
SWMM
NR
10
aJb/:4
U
General
Buildup-
Washoff
MUSLE
Continuous
Subhouriy
yes
EPA/
CEAM
HSPF
NR,RF,
RQ
12
aJbjcA
U,R
General
Loading-
Washoff
Detach-
ment
Continuous
Subhouriy
yes
EPA/
CEAM
CREAMS
NR
15
R
(field
scale)
Potency
Factors
Continuous
Daily
yes
USDA/A
RS
Auto Q-
ILLUDAS
NR
14
a
U
SJN,C,
NC,0
Buildup-
Washoff
Continuous
Event
no
Illinois
SWS
Watershed
Management
Model
RQ
16
a
U,R
N,M
Loading
Function
NA
Annual
no
Florida
DER
Key to References
1. Hydroscience, 1979
2. Driscoll el al., 1990
3. Driver and Tasker, 1988
4. Haith et al., 1992
5. Young et al. 1986
6. HEC, 1977a
7. Beasley & Huggips, 1981
8. Walker et al., 1989
9. Arnold et al., 1991
10. Huber & Dickinson, 1988
11. Dillaha et al., 1988
12. Johanson et al., 1984
13. Alley & Smith, 1982b
14. Terstriep et aL, 1990
15. Knisel, 1980
16. CDM, 1992
Key to Reviews
a. US. EPA, 1992
b. Donigian & Huber, 1991
c WPCF, 1989
d. McKeon & Segna, 1987
3-23
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
3.3.3 Models of Receiving Water Flow
Table 3-3 summarizes a number of models fox the simulation of receiving water
flows. As above, the list focusses on models supported by Federal agencies. This
constraint necessarily omits some of the more complex and interesting model
applications available in the literature; a good summary of such models available for
estuarine applications is provided in Ambrose et al. (1990). The first four columns of
Table 3-3 are similar to those in the pre ous tables.
Col. 5 identifies the major waterbody types for which the flow simulation is
appropriate. These include rivers (i.e., advection dominated systems), lakes/ reservoirs,
and estuaries. Models designated "Lake" are usually appropriate for both lakes and
reservoirs; those intended specifically for reservoirs are indicated as "Reservoir."
Col. 6 shows the dimensionality of the flow simulation. Note that some models
(e.g., DYNHYD) are essentially one-dimensional in their equations, but address multi-
dimensional simulation by using a link-node network.
Col- 7 summarizes the basic method of computing or routing flows. Water
balance and hydrologic routing procedures are based on continuity only, without
solution of the energy equation. Kinematic and dynamic wave methods address time-
dependent momentum at various levels of complexity (see Section 3.2.1). Steady state
solutions to the energy equations are provided by Bernoulli equation type solutions.
CoL 8 identifies whether flows are simulated as steady-state or dynamic, and what
approximate temporal resolution is achieved. The latter is a rather subjective
determination, as a model's ability to accurately represent flows is often at a rather
coarser time step than that needed for internal simulation purposes.
3-24
-------
The Cadmus Group> Inc.
DRAFT
June 8,1993
Table 3-3. Receiving Water Quantity (Flow) Models
Model
Other
Uses
Main
Ref.
Rev.
Waterbody
Type
Dimension-
ality
Routing
Type
Time Step
Agency
QUAL2E
RQ
1
River
Estuary
1-D
Water
Balance
Steady
EPA/
CEAM
DYNHYD (WASP4)
RQ
2
River
Lake
Estuary
1-D+ (to
3D link
node)
Dynamic
Wave
Transient
Subhourly
EPA/
CEAM
AGNPS
NQ,NR,
RQ
3
a,b
River
1-D
Unit
Hydrograph
Transient
Hourly
USDA/ARS
SWRRBWQ
NF,NQ,
RQ
4
a,b
Lake
River
1-D, 0-D
Water
Balance
Transient
Daily
USDA/ARS
HSPF
fJF,NQ,
RQ
5
a,b,c,d
River
Lake
1-D
Hydrologic
Routing
Transient
Hourly
EPA/
CEAM
HEC-2
6
d
River
1-D
Bernoulli
Equation
Steady
HEC
HEC-6
RQ
7
d
River
1-D
Bernoulli
Equation
Series of
steady evts.
HEC
DWOPER
8
d
River
1-D
Dynamic
Wave
Transient
Subhourly
NWS
TR-20
NR
9
River
1-D
Storage +
Kinematic
Transient
Subhourly
SCS
HEC-3
11
Reservoir
1-D
Water
Balance
Monthly
HEC
3-25
-------
The Cadmus Group, Inc. DRAFT June 8,1993
Table 3-3 (continued)
HEC-5
12
Reservoir
River
1-D
network
Hydrologic
Routing
Transient
Hourly
HEC
HEC-1
NF
14
River
Reservoir
1-D
Kinematic &
others
Event
HEC
| CE-QUAL-R1
RQ
15
Reservoir
1-D
thermal &
density
Transient
WES
1 CE-QUAL-W2
RQ
16
Lake
Estuary
2-D
thermal &
density
Transient
WES
| RWQM (STORM)
RQ,NF,
NQ
17
River
1-D
Kinematic
Wave
(to support
longer term
quality)
HEC
I WQRSS
RQ
18
River
Reservoir
1-D
Multiple
methods
Transient
Subhourly
HEC
TABS-2
RQ
19
River
Estuary
2-D
Transient
Subhourly
WES
WIFM-SAL
RQ
20
River
Estuary
2-D
Transient
Subhourly
WES
Key to References
1. Brown & Barnwell, 1987
2. Ambrose et al., 1988
3. Young et al., 1987
4. Arnold et al., 1991
5. Johanson et al., 1984
6. HEC, 1973
7. HEC, 1977b
8. Fread, 1978
9. SCS, 1979
11. HEC, 1976
12. HEQ 1979b
14. HEC, 1985
15. WES, 1988
16. draft available
17. HEQ 1979a
18. HEC, 1978
19. Thomas and McAnally,
1985
20. Schmalz, 1985
Key to Reviews
a. U.S. EPA, 1992
b. Donigian & Huber, 1991
c. WPCF, 1989
d. McKeon & Segna, 1987
3-26
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
3.3.4 Models of Receiving Water Quality
The final table, Table 3-4, summarizes model capabilities for simulation of
receiving water quality. In addition to categories represented in Table 3-3, the following
data are included:
Col. 6 indicates the types of constituents addressed, using the same abbreviations
as in Table 3-2, with some additions; that is, they are characterized as applicable to
nutrients (N), oxygen and oxygen demand (O), metals (M), conservative organic
pollutants (C), nonconservative, reactive organic pollutants (NC), sediment (S), and
temperature (T).
Col. 8 summarizes sediment transport simulation abilities. The options are not
addressed ("NA"), transport of sediments ("transp")/ and simulation of erosion ("eros")
and deposition ("dep").
Col. 10 indicates ability of the pollutant transport routines to simulate degradation
and transformations. Many models incorporate simple decay routines, but cannot
handle transformation products. Those which address both are denoted "full".
3-27
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 3-4. Receiving Water Quality Models
| Model
Other
Uses
Main
Ref.
Rev.
Water-
body type
Consti-
tuents
Dimensi-
onality
Sedi-
ment
Sim.
Time
Step
Transf.
Degrd.
Special
Charac-
teristics
Agncy
QUAL2E
RF
1
River
Estuary
ONAC
NC,T
1-D
NA
Steady
Degr
Uncert.
analysis
EPA/
CEAM
WASP4
RF
2
d
River
Estuary
Lake
General
1-D+ (to
3-D link
node)
transp
Cont.
Subhrly
full
EUTRO&
TOXI
versions
EPA/
CEAM
AGNPS
NONF,
RQ
3
a/b
River
NAC
1-D
transp
eras/
dep
Cont.
Hourly
no
USDA/
ARS
SWRRBWQ
NONF,
RF
4
a,b
River
(sed)
Lake (N)
S,N,C,NC
2-D
Compart
ment
transp
eras.
dep.
Cont.
Daily
Degr
Uncer-
tainty
Analysis
USDA/
ARS
HSPF
NQNF,
RF
5
a,b,
cA
River
Lake
General
1-D
transp
Cont
Subhrly
full
Freq-Dur
module
EPA/
CEAM
EXAMS-II
6
d
General
QNC
1-D to 3-
D
NA
Steady-
Monthly
full
Flows
spec
external
EPA/
CEAM
MEXAMS
7
d
General
M
1-D to 3-
D
NA
Steady-
Monthly
full
Flows
spec
external
EPA/
CEAM
HEC-6
RF
8
d
River
Reservoir
S
1-D
transp
eros
dep
Contin
Daily
NA
stream
bed
profiles
HEC
3-28
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
V
Table 3-4 (continued)
STREAMDO
IV
10
River
O
1-D
NA
Steady
NA
ammonia
tox.
EPA R
VIII
AMMTOX
11
River
Ammonia
1-D
NA
Monthly
Diel
cycles
Degr
Freq
Analysis
EPA R
Vffl
SMPTOX3
12
River
QNC
1-D
NA
Steady
Degr
PS only
Sens.
Anal.
EPA/
CEAM
FHWA
NF,NQ
13
a4>
River
Lake
N,C,M
Statistical
NA
Stats on
Events
no
Highway
Runoff
FHWA
Watershed
Management
8 Model
NQ
14
a
Lake
N,M
0-D
NA
Annual
simple
Degr
Spread-
sheet
Ha.
DER
I DYNTOX
17
River
C,NC
(PS +
backgrd)
1-D
NA
Daily
Degr
Freq- Dur
analysis
EPA/
CEAM
SEDDEP
18
Estuary
S
2-D
trans
depp
NA
outfall
depos.
EPA/
CEAM
CE-QUAL-R1
RF
19
Reservoir
T,C,NC,
N,0
1-D
transp
dep
Hourly
full
Monte
Carlo
WES
CE-QUAL-
W2
RF
20
Lake
Estuary
T,CNC,
N,0
2-D
transp
Hourly
full
full
energy
WES
RWQM
(STORM)
NF^a
RF
21
River
0,N,T
1-D
NA
long-
term
Degr
Linked to
STORM
HEC
3-29
-------
The Cadmus Group; Inc.
DRAFT
June 8,1993
Table 3-4 (continued)
WQRSS
RF
22
River
Reservoir
General
1-D
transp
Qmtin.
Subhiiy
full
temp.,
ecologic
sim.
HEC
CORMIX
23
Estuary
Lake
Dilution
2-D, 3-D
NA
Steady
no
near field
EPA/
CEAM
SEM
24
Estuary
0,C,NC
1-D
NA
Steady
degr
flows
input
EPA
AUTOQUAL
25,26
River
Estuary
Q,N
1-D
NA
Steady
no
flows
input
EPA
FETRA
27
d
Estuary
Lake
S,C,NC
2-D
transp
eros
dep
Gontin.
degr
finite
element
NUREG
TABS-2
RF
28
River
Estuary
S,C,NC
2-D
transp
dep
Contin.
degr
WES
WIFM-SAL
RF
29
River
Estuary
bacteria
2-D
NA
Gontin.
Subhrly
degr
WES
PLUMES
30
Estuary
Lake
Dilution
2-D, 3-D
NA
Steady
no
near field
EPA/
CEAM
Key to References
lU
zanaer & Love, iw
z*. bchmalz, ivk>
1. Brown & Barnwell, 1987
11.
Saunders et al., 1993
21. HEC, 1979a
30. CEAM BBS
2. Ambrose et al., 1988
12.
LimnoTech, 1993
22. HEC, 1978
3. Young el al., 1986
13.
Driscoll et al., 1990
23. Jirka, 1992
Key to Reviews
4. Arnold et al., 1991
14.
CDM, 1992
24. Hydrosdence, 1971
a. U.S. EPA, 1992
5. Johanson et al., 1984
17.
LimnoTech, 1985
25. Lovelace, 1975
b. Donigian & Huber, 1991
6. Bums & dine, 1985
18.
Bodeen et al., 1989
26. Crim & Lovelace, 1973
c. WPCF, 1989
7. Bums et al., 1982
19.
WES, 1986
27. Onishi, 1981
d. McKeon & Segna, 1987
8. HEC, 1977b
2a
draft available
28. Thomas & McAnally, 1985
3-30
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
References. Chapter 3
Alley, W.M. and P.E. Smith. 1982a. Distributed Routing Rainfall-Runoff Model, Version
II. Open-File Report 82-344. USGS, Denver, CO.
Alley, W.M. and P.E. Smith. 1982b. Multi-Event Urban Runoff Quality Model. Open-File
Report 82-764. USGS, Denver, CO
Ambrose, R.A., T.A. Wool, J.P. Connolly and R.W. Schanz. 1988. WASP4, A
Hydrodynamic and Water Quality Model - Model Theory, User's Manual and
Programmer's Guide. EPA/600/3-87/039. ERL Athens.
Arnold, J.G., J.R. Williams, R.H. Griggs and N.B. Sammons. 1991. SWRRBWQ, A Basin
Scale Model for Assessing Management Impacts on Water Quality. USDA
Agricultural Research Service, Grassland, Soil and Water Research Laboratory,
Temple, TX
Beasley, D.B. and L.F. Huggins. 1981. ANSWERS Users Manual. EPA-905/9-82-001.
U.S. EPA, Region V, Chicago, IL
Bodeen, C.A., T.J. Hendricks, W.E. Frick, D.J. Baumgartner, J.E. Yerxa and A. Steele.
1989. User's Guide for SEDEP: A Program for Computing Seabed Deposition
Rates of Outfall Particulates in Coastal Marine Environments. EPA ERL Athens,
report 109-ERL-N.
Brown, L.C. and T.O. BarnwelL 1987. The Enhanced Stream Water Quality Models
QUAL2E and QUAL2E-UNCAS. EPA-600-3-87-007. NTIS PB87-202156.
Burns, L.A. and D.M. Cline. 1985. EXAMS-H: Exposure Analysis Modeling System,
Reference Manual for EXAMS-H. EPA-600/3-85-038. ERL Athens, GA
Burns, L.A., D.M. dine and R.R. Lassiter. 1982. Exposure Analysis Modeling System
(EXAMS): User Manual and System Documentation. EPA-600/3-82-023. ERL
Athens, GA
CDM. 1992. Watershed Management Model User's Manual, Version 2.0. Florida Dept.
of Environmental Regulation, Tallahassee, FL
Crim, R. and NX. Lovelace. 1973. AUTO-QUAL Modeling System. EPA-440/9-73-004.
3-31
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Dillaha, T.A., C.D. Heatwole, M.R. Bennett, S. Mostaghimi, V.O. Shanholtz and B.B.
Ross. 1988. Water Quality Modeling for Nonpoint Source Pollution Control
Planning: Nutrient Transport. Report No. SW-88-02. Dept. of Agricultural
Engineering, Virginia Polytechnic Institute and State University.
Donigian, A.S. Jr. and W.C. Huber. 1991. Modeling of Nonpoint Source Water Quality
in Urban and Non-urban Areas. EPA/600/3-91/039. Environmental Research
Laboratory, Athens, GA
Driscoll, E.D., P.E. Shelley and E.W. Strecker. Pollutant Loadings and Impacts from
Highway Stormwater Runoff, Volume I: Design Procedure. NHS PB90-257551.
Federal Highway Administration, McLean, VA
Driver, N.E. and G.D. Tasker. 1988. Techniques for Estimation of Storm-Runoff Loads,
Volumes, and Selected Constituent Concentrations in Urban Watersheds in the
United States. Open-File Report 88-191. USGS, Denver, CO
Fread, D.L. 1978. National Weather Service Operational Dynamic Wave Model. National
Oceanic and Atmospheric Administration, Washington, DC
Freedman, P.L., D.W. Dilks and B.A. Monson. 1992. Technical Guidance Manual for
Performing Waste Load Allocations, Book HI: Estuaries, Part 4: Critical Review of
Coastal Embayment and Estuarine Waste Load Allocation Modeling. EPA-823-R-
92-005. Office of Water (WH-585).
Haith, D.A., R. Mandel and R.S. Wu. 1992. GWLF, Generalized Watershed Loading
Functions, Version 2.0, User's Manual. Dept of Agricultural & Biological
Engineering, Cornell University, Ithaca, NY
HEC. 1979a. Receiving Water Quality Model, Program User's Manual. Hydrologic
Engineering Center, Corps of Engineers, U.S. Army, Davis, CA
HEC. 1979b. HEC-5 Simulation of Flood Control and Conservation Systems, Program
User's Manual. Hydrologic Engineering Center, Corps of Engineers, U.S. Army,
Davis, CA
HEC. 1985. HEC-1 Flood Hydrograph Package, User's Manual. Hydrologic Engineering
Center, Corps of Engineers, U.S. Army, Davis, CA
HEC. 1978. Water Quality for River-Reservoir Systems, Program User's Manual.
Hydrologic Engineering Center, Corps of Engineer's, U.S. Army, Davis, CA
3-32
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
HEC. 1977a. Storage, Treatment, Overflow, Runoff Model "STORM", User's Manual.
Hydrologic Engineering Center, Corps of Engineers, U.S. Army, Davis, CA
HEC. 1977b. HEC-6 Scour and Deposition in Rivers and Reservoirs, Program User's
Manual. Hydrologic Engineering Center, Corps of Engineers, U.S. Army, Davis,
CA
HEC. 1976. HEC-3 Reservoir System Analysis for Conservation, Program User's Manual.
Hydrologic Engineering Center, Corps of Engineers, U.S. Army, Davis, CA
Huber, W.C. and R.E. Dickinson. 1988. Storm Water Management Model Version 4,
User's Manual. EPA/600/3-88/001a (NTIS PB88-236641/AS). Environmental
Research Laboratory, Athens, GA
Hydroscience. 1979. A Statistical Method for Assessment of Urban Storm Water Loads-
Impacts-Controls. EPA-440/3-79-023 (NTIS PB-299185/9)
Hydroscience. 1971. Simplified Mathematical Modeling of Water Quality. EPA 1971-44-
367/392.
Istok, J. 1989. Groundwater Modeling by the Finite Element Method. Water Resources
Monograph 13. American Geophysical Union, Washington, DC
Jirka, G.H. 1992. Technical Guidance Manual for Performing Waste Load Allocations,
Book ni: Estuaries, Part 3: Use of Mixing Zone Models in Estuarine Waste Load
Allocations. EPA-823-R-92-004.
Johanson, R.C., J.C. Imhoff, J.L. Kittle, Jr. and A.S. Donigian. 1984. Hydrological
Simulation Program - FORTRAN (HSPF), Users Manual for Release 8.0. EPA-
600/3-84-066. Environmental Research Laboratory, Athens, GA
Knisel, W. G. 1980. CREAMS, A Field Scale Model for Chemicals, Runoff, and Erosion
from Agricultural Management Systems. USDA Conservation Research, Report
No. 26
LimnoTech. 1993. Simplified Method Progjam - Variable»Complexity Stream Toxics
Model (SMPTOX3), Version 2.01. Available from
Lovelace, N.L. 1975. AUTO-QUAL Modeling System: Supplement 1: Modification for
Non-Point Source Loadings. EPA-440/9-73-004.
3-33
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
Lyman, W.J., W.F. Reehl and D.H. Rosenblatt. 1990. Handbook of Chemical Property
Estimation Methods, Environmental Behavior of Organic Compounds. American
Chemical Society, Washington, DC
McKeon, T.J. and J.J. Segna. 1987. Selection Criteria for Mathematical Models Used in
Exposure Assessments: Surface Water Models. EPA/600/8-87/042. Exposure
Assessment Group, Office of Health and Environmental Assessment, U.S. EPA,
July 1987.
Mills, W.B. et al. 1985. Water Quality Assessment: A Screening Procedure for Toxic and
Conventional Pollutants in Surface and Ground Water (Revised 1985).
EPA/600/6-85/002a-b. Environmental Research Laboratory, Athens, GA
Onishi, Y. and S.E. Wise. 1982. User's Manual for the Instream Sediment-Contaminant
Transport Model SERATRA. EPA-600/3-82-055.
Roesner, L.A., J.A. Aldrich and R.E. Dickinson. 1988. Storm Water Management Model
User's Manual, Version 4, EXTRAN Addendum. EPA/600/3-88/001b (NTIS
PB84-198431). Environmental Research Laboratory, Athens, GA
Saunders, J.F., W. M. Lewis and A. Sjodin. 1993. Ammonia Toxicity Model AMMTOX
Version 1.0, Operator's Manual. (Distributed by EPA Region VIE, Denver, CO)
Schmalz, R.A. 1985. User Guide for WIFM-SAL: A Two-Dimensional Vertically
Integrated, Time-Varying Estuarine Transport Model. Waterways Experiment
Station, Vicksburg, MS
SCS. 1986. Urban Hydrology for Small Watersheds. Technical Release #55 (2nd edition)
SCS. 1979. Simplified Dam-Breach Routing Procedure. Technical Release 66, U.S. Dept.
of Agriculture, Soil Conservation Service.
SCS. 1973. Computer Program for Project Formulation Hydrology. Technical Release #20
Terstriep, M.L., M.T. Lee, E.P. Mills, A.V. Greene and M.R. Rahman. 1990. Simulation
of Urban Runoff and Pollutant Loading from the Greater Lake Calumet Area.
Prepared by the Illinois State Water Survey for the U.S. Environmental Protection
Agency, Region V, Water Division, Watershed Management Unit, Chicago, IL
3-34
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Thomas, W.A. and W.H. McAnally. 1985. User's Manual for the Generalized Computer
Program System - Open Channel Flow and Sedimentation TABS-2. Waterways
Experiment Station, Vicksburg, MS
U.S. EPA. 1992. Compendium of Watershed-Scale Models for TMDL Development.
EPA/841-R-92-002. U.S. EPA Office of Wetlands, Oceans and Watersheds and
Office of Science and Technology.
Walker, J.F., S.A. Pickard and W.C. Sonzogni. 1989. Spreadsheet watershed modeling for
nonpoint-source pollution management in a Wisconsin basin. Water Resources
Bulletin 25(1): 139-147.
WPCF. 1989. Combined Sewer Overflow Pollution Abatement. Manual of Practice No.
FD-17. Water Pollution Control Federation, Alexandria, VA.
Young, R.A., C.A. Onstad, D.D. Bosch and W.P. Anderson. 1986. Agricultural nonpoint
source pollution model: A watershed analysis tooL U.S. Dept. of Agriculture,
Agricultural Research Service, Morris, MN
Zander, B. and J. Love. 1990. STREAMDO IV and Supplemental Ammonia Toxicity
Models. U.S. EPA Region VIE, Denver, CO
3-35
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
Chapter IV. TMDL Modeling Strategy and Model Identification
Purpose: This chapter is intended to guide the user in formulating an appropriate
modeling strategy for establishing a TMDL. Key aspects of the modeling strategy are
the identification and selection of an appropriate model or models. Following Ambrose
et al. (1990), the goals of model identification and selection are to identify the simplest
conceptual model that includes the physical and chemical phenomena that should be
addressed in establishing the TMDL, and to select the most useful analytical formula,
statistical model, or computer simulation model for calculating the TMDL.
We first discuss the principles of establishing a modeling strategy. This is
followed by a discussion of general criteria for model identification for TMDLs, and
specific considerations for TMDLs involving wet-weather PS and NPS loads. In Section
4.4 we have endeavored to apply these principles to develop practical guidance for
model selection. This is presented in the form of a set of structured decision criteria, or
decision trees.
The purpose of the decision trees is to assist in identifying an optimal, site-specific
modeling approach, taking into consideration the characteristics of the loads,
characteristics of the receiving waterbody, data availability, regulatory context, and
available resources for analysis. This is done by leading the user step-by-step through
a hierarchy of issues which should be considered in the model identification process.
The modeling decision trees are primarily concerned with model identification, rather
than selection; they are not an 'expert system' that will pick which model to use. Rather,
the presentation consists of a logically structured summary of suggestions and guidance
to the type of models appropriate to a specific TMDL situation, and the corresponding
monitoring requirements. The output of the process is a statement of necessary model
criteria for a given TMDL. The user can then match these criteria to the characteristics
of available models to complete the model selection process. Where no match is found
this may indicate the need to develop a site-specific model.
4.1 Establishing a Modeling Strategy
In the TMDL process, modeling may- be used to accomplish several related
functions. First, modeling, combined with monitoring information, can be used to
establish the total assimilative capacity of the waterbody. Modeling, combined with
monitoring, may also be used to estimate the frequency distribution of loads from wet-
weather episodic loads and any resulting excursions of WQSs. Finally, modeling often
provides the basis for the evaluation of BMPs and optimal assignment of LAs and
WLAs. The quality of the TMDL may thus be critically dependent on implementation
4-1
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
of an appropriate modeling strategy.
Of course, not all TMDLs require modeling; in some cases decisions can be made
directly from observational data. Even when models are needed, it is not always
necessary to employ a detailed simulation model if simple scoping models are adequate
(Figure 4-1). However, most TMDLs will require some form of modeling. The type of
modeling employed will be closely linked with the availability of data. While models
are often used to extrapolate beyond m sured data, the type of model that can be
employed is also constrained by the availability of data for calibration and verification.
EPA's Science Advisory Board (U.S. EPA 1989) has concluded that, ideally, modeling
should be linked with monitoring data in regulatory assessments. Thus the modeling
and monitoring strategies for TMDLs should be developed in tandem. Further, the
modeling strategy will likely need to be periodically refined as additional data are
collected and initial modeling efforts applied.
A complete modeling strategy has a number of components, which range from
setting objectives to the identification, selection and implementation of appropriate
models. Many of these aspects can be thought of as analogous to the QA/QC measures
which are applied to measurements (see 57 Federal Register 22907, May 29,1992). The
modeling strategy has four major components:
4.1.1. Modeling Study Objectives
The first step in development of the modeling strategy is to clearly define the goal
of the modeling exercise, i.e. how a model can help address the questions and problems
presented by the TMDL. Note that the modeling study objectives are likely to change
as additional data are developed, particularly when a Phased TMDL is pursued.
Modeling study objectives should include a clear statement of what information
the model will help estimate, and how this estimate will be used. For TMDL or WLA
analysis, the objectives will address WQSs or beneficial uses. Therefore, the first step
is to review the applicable WQSs and uses to be protected. Local, state, and federal
regulations may contribute to a set of objectives and constraints. Each may specify
particular pollutants or classes of pollutants, and imply time and space scales that must
be resolved by the model. For example, proscription of "toxic pollutants in toxic
amounts" implies simulation of whole effluent toxicity dilution. Ammonia or metals
standards imply simulation of those specific chemicals (Ambrose et al., 1990). The
modeling study objectives also must be consistent with known project constraints (i.e.,
schedule, budget, and other resources), as well as the objectives of the receiving water
analysis.
4-2
-------
Figure 4-1. Basic Conceptual Decision Tree For Model Choice
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
The modeling study objectives provide the basis for model identification, and are
developed further in Section 4.4.
4*1.2. Model Identification and Selection
The second component of a modeling strategy is the identification and selection
of appropriate models (whether simple or complex). This is the focus of Section 4.4.
4.1.3. Model Data Collection Plan
Modeling and monitoring efforts will typically develop in tandem: In most cases
it will not be possible to specify all the necessary components of the modeling strategy
at the beginning of the process, when data may be limited. Instead, it will be necessary
to refine the goals and requirements of the modeling as additional data are collected.
Similarly, the monitoring plan may need to be adjusted based on preliminary modeling
runs. For instance, in a DO/BOD problem involving steady point and episodic sources,
the initial stage of analysis, with limited data in hand, might consist of steady state
stream modeling covering a number of different scenarios for the impacts of the steady
plus episodic loads under a variety of streamflow conditions. This analysis would reveal
the general location of critical points - although the identification will likely be much less
exact than in a problem involving only steady point loads. This initial analysis would
yield only a relative, and not quantitative estimate of the occurrence and probable
location of WQS excursions. It would also yield an indication of which episodic loads
are likely to be of most significance to the overall impairment of the waterbody. This
initial, simple modeling analysis could provide the basis for a targeted monitoring
strategy, designed to further quantify those loads which might be important contributors
to waterbody impairment. The additional data collected by this targeted monitoring
program could then form the basis for a dynamic (unsteady) modeling application.
These observations indicate that the development of a final modeling strategy is
an iterative process, involving close interplay between the modeling and monitoring
plans. Development of monitoring plans is discussed further in Chapter V.
4.1.4. Model Quality Assurance
Just as quality assurance (QA) principles are applied to the collection of analytical
data, analogous QA measures can be applied to simulation modeling. Quality assurance
consists of a plan to assure that a product meets defined standards of quality with a
stated level of confidence. Once a modeling approach is identified, a QA approach
should be developed to provide guidelines for the project and establish a baseline
4-4
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
against which the effectiveness of model applications can be evaluated.
As with analytical QA, modeling quality assurance begins with the establishment
of data quality objectives (DQOs). The DQOs should reflect the modeling study
objectives, and provide a pre-established benchmark against which to evaluate the
success and validity of the modeling effort. For instance, it is advisable to establish a
priori goals for the desired accuracy of model predictions, goodness of fit for model
calibration and verification, as well as sampling uncertainty in the determination of
measurable physical parameters. Such goals should be defined in terms of the type of
model used,, and the objectives of the modeling study. For instance, the DQOs for a
preliminary study concerned with average nutrient levels in a waterbody would be quite
different from those for a model required to predict point-in-time coliform concentrations
at a public beach. In the first case, a simple model which gave an accurate prediction
of seasonal averages is likely to be quite sufficient for estimating a TMDL; in the latter
case it may be necessary to conduct more detailed modeling sufficient to determine the
time history of waterbody impacts from individual loading events.
Of course, the relative success and accuracy of most environmental fate and
transport model applications is more difficult to predict and control than the analysis of
chemical samples. When model DQOs are not met, this should certainly prompt a
reexamination of model objectives, model application, and data collection. However,
failure to attain desired levels of accuracy does not mean that the model results are
unusable (unlike laboratory QA). Instead, it may imply that (1) the modeling strategy
should be revised, or (2) the model results should be used to estimate the TMDL, but
with a correspondingly large MOS to account for the potential error in model
predictions, limit excursions of WQSs, and protect beneficial uses. Evaluation of model
accuracy, uncertainty and bias is covered in Chapter 7.
To further the analogy with laboratory QA, the user may consider formalizing a
quality control method as well. Quality control (QC) is a part of QA, and ensures the
product of the analysis is satisfactory and economical through the specification of and
adherence to certain procedures and protocols. Quality control procedures for modeling
may include testing the installed version of computer code for correct performance,
validating accuracy of input data sets, good record keeping and documentation of model
calibration, verification and application, and so on.
4.2 General Criteria for a Modeling Strategy for TMDLs
This section addresses the general question of determining a modeling strategy
for calculating TMDLs. Formulation of such a strategy depends on many factors. It is
more than just choosing the "best" model of a physical system, as the strategy must take
4-5
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
into account such real-world constraints as available resources for a project. This
guidance takes a somewhat different approach from some of EPA's Waste Load
Allocation modeling guidance in that it recognizes the inevitability of such constraints
and attempts to define an appropriate level of analysis to perform a TMDL with finite
resources.
Models used to calculate TMDLs must first adequately represent the significant
features of the physical system describf . Passing this test, a key aspect of model
selection is the cost and effort required for implementation, which must be balanced
against the benefits achieved by use of the model. There is an obvious tension between
an optimal technical representation of the physical system and the cost of
implementation, as increasing model resolution (the fineness of the spatial and temporal
scales of model prediction) and model accuracy (the extent to which predictions differ
from observations) usually involves increased expense and effort. Further, complete
accuracy can never be achieved (especially for simulations involving wet-weather PS and
NPS loading), and there is typically a point at which increased modeling effort provides
rapidly diminishing returns in terms of increased accuracy. Also, as noted by Thomann
in Freedman et al. (1992), increasing model complexity beyond a certain point actually
results in a decline in model credibility (Figure 4-2). This is because increasing
complexity generally requires specification of more and more parameters and state
variables, all of which require a detailed database for complete assessment. Indeed,
attempts to develop models with extremely high levels of accuracy and resolution can
have the unintended effect of delaying the analysis and implementation of controls that
may be necessary to prevent impairment of the waterbody. On the other hand, simple
modeling approaches which offer only approximate representations of the system can
often be used to implement TMDLs. Ideally, the effects of the inaccuracies and
approximations introduced by use of the simpler approach can be explicitly incorporated
into the MOS. At a minimum, the simple analysis should yield an indication of the
relative risk posed by different sources. This will allow the TMDL developer to target
the most promising sources for control during a phased TMDL development process.
An appropriate modeling strategy must be focused on performing the TMDL in a
practical manner. It therefore involves more than selecting modeling techniques that
provide an optimal description of the physical system and pollutants or impact
mechanisms of interest. In general, we advise designing a modeling effort to provide
answers that are as detailed and accurate as needed, at the lowest corresponding expense
and effort - e.g., don't run a complex simulation with HSPF if a spreadsheet dissolved-
oxygen simulation will be adequate to estimate the WLAs and LAs. In EPA's estuarine
WLA guidance (Freedman et al., 1992), this point of view is succinctly summarized by
Thomann:
The best models are often the simplest...[advocate) doing estuarine water quality
modeling in as simple a fashion as possible and only after all simplicity has been
4-6
-------
Level of Model Complexity
Level of Model Complexity
Figure 4-2. Relation of Model Complexity,
Accuracy, and Credibility
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
exhausted should increasing complexity be introduced and then only after careful
consideration is given to the improvements in the model that might be realized.
The reasons for this bias are: (a) most analysts have only limited experience, time
and resources available, and (b) unnecessarily complex models sometimes tend
to obscure uncertainty behind a facade of "reality".
Defining the optimum balance between ease-of-implementation and accuracy is always
difficult, yet is a key issue in model identification.
4.2.1 Types of Criteria for Model Identification
While the physical characteristics of the system certainly must guide selection of
models, other factors must also be taken into account The criteria for model
identification for TMDLs fit into three general categories (expanding on the classification
of Mao, 1992): Technical Criteria, Regulatory Criteria, and User (Functional &
Operational) Criteria.
Technical Criteria comprise the match of the model to the physical characteristics
of the system. They reflect whether the model is appropriate to the physical system
being described. For instance, a one-dimensional transport model designed for rivers
cannot provide a good description for the distribution of contaminants in a stratified
estuary, and slug releases of reactive chemicals cannot be accurately described by steady
state models. These considerations first involve the comparatively simple question of
whether a model's governing equations and boundary conditions are a good match to
the characteristics of the waterbody, washoff. process, and pollutants requiring TMDLs.
However, evaluation relevant to the TMDL process involves additional, and potentially
competing considerations: For instance, as shown in the simple example above, it is
possible to perform a TMDL that is protective of WQSs using a model which does not
provide a truly accurate description of the system (e.g., no dispersion). Technical criteria
must then 1) recognize the differences between the model and reality, and 2) insure that
all such discrepancies result in errors on the side of safety.
Regulatory Criteria reflect the fact that the modeling effort is driven by
compliance with the CWA, and should be framed in the appropriate regulatory context
That is, the TMDL is based on attaining water quality criteria and die modeling effort
should provide answers framed in similar terms. Determining whether a numeric
criterion will be achieved may require a different strategy than would be required for
a narrative criterion. Further, modeling for a numeric criterion that specifies an average
concentration over a large area will require a different level of detail than for the
evaluation of a numeric criterion based on point concentrations in time.
User Criteria comprise the functional and operational needs of the user.
4-8
-------
The Cadmus Group, Jnc,
DRAFT
June 8, 1993
Assuming a variety of models are available to provide an adequate physical description
of the system, and to provide answers to the appropriate regulatory questions, then
selection among the candidate models will involve consideration of such things as
available resources, ease of use, and so on. Functional needs refer to such issues as ease
of use and communication of results, availability and adequacy of documentation, and
model complexity and data requirements. The level of effort required to couple
particular washoff and receiving water models can be an important functional criterion.
Use of a model that is poorly documented and not well known will increase the
difficulty of communicating and gaining acceptance of the results. Another key issue
is the interface between modeling and monitoring. A complex model implementation
is of little value unless adequate data are available for calibration and verification. In
many cases, the best approach may be an incremental effort, in which simple screening
models are used with initial, usually limited, data. These results may then suggest a
more detailed modeling and monitoring strategy. Implementation of the more detailed
model will then require additional refinements in monitoring for calibration/verification,
and so on. Operational needs reflect both the requisite technical ability to implement
the model, and the estimate of cost and time requirements for the implementation.
These criteria provide the cost side for a cost-benefit analysis of model selection. In
general, use of more complex models can result in significantly higher costs, due both
to implementation effort and amount of data required for model calibration. This cost
must be balanced with available resources, and the benefits to be achieved through the
effort. The benefits of modeling are defined through 1) additional protection against
excursions of WQSs, and 2) minimization of the economic loss incurred by requiring the
maintenance of an excess MOS to compensate for uncertainty in predictions regarding
the system. In many instances, lower cost, simpler modeling approaches will often be
available which will enable completion of a TMDL, given that an adequate MOS is
calculated to account for the simplification. Greater modeling effort can then result in
the benefit of a reduced MOS required to protect designated uses (and thus a lesser
burden on PS and NPS dischargers), but at the cost of increased resource requirements
for modeling.
4.2.2 Idealized Approach to Modeling Strategy
As noted in the previous section, there are three general sets of considerations or
modules that control the determination of a modeling strategy: 1) technical criteria
(physical characteristics of the chemical and system); 2) regulatory criteria; and 3) user
criteria (functional and operational needs). These three modules should interact as a
three-dimensional matrix to determine the optimal modeling strategy (see Figure 4-3).
In theory, one could evaluate all three modules simultaneously, leading to the
determination of an optimal modeling strategy. However, simultaneous evaluation,
4-9
-------
Regulatory Criteria
Figure 4-3. Determining an Optimal Modeling
Strategy for the TMDL
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
while it could be addressed in an artificial intelligence context, is difficult to formulate
into written guidance. It is also desirable to separate the determination of physical
characteristics from the other modules, as it may be necessary to address multiple
chemicals, as well as differing levels of regulatory criteria and functional needs for a
single set of physical characteristics. (Noting, however, that some physical characteristics
of the receiving water may also change in response to changes in loads.)
An idealized approach, which might be followed by an expert with detailed
knowledge of a specific waterbody, is shown in Figure 4-4. This represents a "top down"
approach, in which a thorough, pre-existing understanding of the site is used to select
an appropriate modeling strategy. (This will be contrasted with a practical approach in
4.2.3.)
The top box in Figure 4-4 addresses the determination of physical characteristics,
and also shows the functional relationships of some of the major subcomponents within
this box. The assessment of the physical characteristics (of the chemical, receiving water,
episodic source loads, etc.) forms die first module, and yields the necessary evidence to
formulate an optimal descriptive model strategy for the system - i.e., the model strategy
which might be chosen to advance scientific understanding of die behavior of the
pollutant in the system if time, money, and the particular form of the applicable
regulatory criteria were not of concern.
This module identifies important physical characteristics of chemicals, sources and
hydraulics relevant to modeling. It has a primary aim of describing the level of detail
appropriate to simulations. This in turn depends on characteristics of the chemical (e.g.,
reactive or conservative), of the waterbody (e.g., complexity of flow) and loading (e.g.,
constant vs. wet-weather episodic). The basic strategy is shown in the top box of Figure
4-4. We begin by identifying important characteristics of the chemical(s) under study
as well as hydraulic behavior of the receiving water body. The degree of spatial and
temporal detail required in a receiving water model is determined as a function of both
chemical and hydraulic characteristics. For instance, if we are studying impacts of
nutrient loadings where the response time of the waterbody is relatively slow it may not
be necessary to study short time or small spatial variability in nutrient concentrations
in the waterbody. In other words, we might make do with a description of average
transport processes, at a time scale appropriate to the response of the waterbody, rather
than requiring a minute-by-minute dynamic simulation.
Modeling of wet-weather loads is, of course, partially dependent on pollutant
characteristics. However, the spatial and temporal scale for wet-weather load modeling
will also be driven by the requirements of the receiving water model. That is, if the
receiving water model needs daily loads only, it may not be necessary to simulate hourly
loads, unless this is the best way to get at daily totals.
4-11
-------
Technical Criteria
Chemical
Receiving Water
_>
i
—-——a
f
NPS Loading
Space-Time Discretization
<—
Receiving Water
Space-Time Discretization
Appropriate Descriptive Model
Optimal Modeling Strategy
Figure 4-4. Idealized Approach to TMDL Modeling Strategy
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
In the second module, the descriptive modeling strategy is filtered through the
applicable regulatory criteria. Here the modeling strategy is tuned to provide an
efficient way to answer specific regulatory concerns. Combining the requirements of the
descriptive model with the filter of regulatory criteria results in the specification of an
optimal decision model strategy. This is the strategy which might be chosen if both the
physical characteristics and regulatory criteria were taken into account, but issues such
as cost, availability of models and so on were still not included.
In the third module, the decision modeling strategy is filtered through these
functional and operational needs. This results in the formulation of a practical modeling
strategy which can be applied to the site. Derived in this way, such a strategy attempts
to honor the physical characteristics, and the regulatory decision criteria, while also
seeking an optimal expenditure of money and effort to reach a target level of precision
in the answers.
When substantial expert knowledge of the system is already available, the
idealized top-down approach to modeling strategy shown in Figure 4-4 could be applied.
However, a practical and generalized approach to the selection strategy, applicable to
a wide range of sites, will proceed somewhat differently. This should recognize that the
TMDL analysis typically begins before perfect knowledge of the characteristics of title site
is attained, and that die analysis may proceed in phases, from simple to complex. The
idealized procedure of Figure 4-4 must be rearranged in terms of specific strategic
criteria, which interact with one another and with various components of the three
modules.
The real-world, practical problem of model selection with limited resources and
incomplete data may be rather different from the idealized, theoretical process
schematized in Figure 4-4. A dichotomy often exists between the two viewpoints, which
can result in technical modeling guidance (expressing what ought to be done, given
unlimited resources) being perceived as of little relevance to the solution of real-world
problems (where practical constraints on resources are a primary determinant of
modeling strategy). These practical constraints, which come under the heading of
Functional and Operational Needs, assume up-front significance in the practical
approach to model selection.
How can these two viewpoints be reconciled in addressing the problem of TMDLs
with episodic, wet-weather PS/NPS loading? In general, the most accurate mathematical
description of wet-weather PS/NPS loading to a receiving water would require use of
sophisticated, time-varying simulation programs, requiring a high level of expertise and
considerable expenditure of time and money. However, the objective of the TMDL is
the achievement of WQSs and protection of designated uses, to which accuracy is a tool,
4-13
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
rather than an end to itself. This means that a less accurate, but more readily
implemented, steady state representation of episodic loading can often be used to
provide at least a preliminary assessment, given the determination of an adequate MOS.
As discussed in Section 4.2.1, there is generally a tradeoff between increased accuracy
(and smaller required MOS) and cost/resources required to calculate the TMDL.
In practical terms, most every TMDL will commence with the use of simple
steady-state mass balance models for scoping the problem (unless sophisticated model
representations of the water body are already available and calibrated). This can be
called a Level 0 representation (see Table 4-1). A Level 0 representation is relatively low
cost, but cannot represent time variability of the wet-weather loading, or frequency of
point concentrations, and therefore will require a relatively high MOS (often
accomplished through the imposition of worst-case assumptions).
If the Level 0 representation is unsatisfactory, because the uncertainties in the
analysis or the required MOS are too high, additional, more sophisticated effort must be
brought to bear on the analysis. For instance, the mass balance scoping could be
replaced with a steady-state simulation analysis of the receiving water body. This would
help to refine the estimation of point concentrations in the receiving water, but would
still require use of worst case assumptions to provide an adequate MOS for episodic
wet-weather loading (e.g., maximum probable event loading combined with antecedent
drought conditions in a receiving stream).
We can conceive of the process of moving from simple, lower cost representations
to more complex, higher cost representations as a ladder, in which we start from Level
0, and climb up to higher levete. How far is it necessary to climb? This must be
determined on the basis of the tradeoff between cost (and available resources) and
accuracy (and required MOS). The exact specification of the components of this ladder
will vary from case to case. However, the general concepts remain the same. A typical
example is provided in Table 4-1. This could represent the succession of Levels for a
problem involving conventional pollutants, such as BOD, where response of the
waterbody is spread over space and time. For a decaying constituent with acute toxic
effects the Levels would likely be defined somewhat differently, with greater emphasis
on the time variability of the loading.
Corresponding to the example specifications ef Table 4-1, it-will be necessary to specify
the spatial and temporal resolution of the simulation of the wet-weather loadings, the
receiving water hydraulics, and the chemical transport and reactions. These will depend
on the characteristics of the system and pollutant. For example, the types of description
of receiving water body hydrodynamics which might correspond to the levels in Table
4-1, by waterbody type, are shown in Table 4-2.
4-14
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
Table 4-1. Example Levels of Analysis for TMDLs with Episodic NPS Loading
and Receiving Water Models (Conventional Pollutant)
Level of
Analysis
NPS
Loading
Receiving
Water
Cost
Required
MOS
Data
Require-
ments
Level 0
(Scoping)
Steady/
Generic or
RM
Mass
Balance/
Worst Case
Low
Very
Large
Low
Level 1
Steady/
Average or
RM
Steady
Low
Large
Low
Level 2
Seasonal/
Average or
RM
Seasonal
Low-
Moderate
Large
Low -
Moderate
Level 3
Seasonal/
Event Mean
Daily
Moderate
Moderate
Moderate
Level 4
Event
Dynamic
High
Small
High
In sum, an "ideal" modeling strategy must reflect the physical characteristics of
the site and pollutant, yet also reflect available resources, prioritization, state of phased
TMDL analysis, and acceptable levels of uncertainty. This presents a complex set of
interlocking goals, the resolution of which will vary from site to site- Specifying the
complete criteria for an ideal modeling strategy is not possible a priori, as each receiving
water body is likely to exhibit its own peculiarities. However, it is possible to construct
logically ordered suggestions to assist in the proper formulation of such a strategy,
recognizing that the picture must be completed with site-specific knowledge.
The structured decision process is presented in Section 4.4 as a sequence of
logically constructed questionnaires and decision trees. These are designed to help the
user develop a set of formal model selection criteria, which may then be matched to the
available models, such as those described in Chapter III.
4-15
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 4-2. Example Levels of Analysis for Receiving Water Models
Level of Analysis
River
Lake
Estuary
Level 0
Mass Balance
Mass Balance
Not Applicable
Level 1
1-D Steady
1-D/ 2-D Steady
1-D/ 2-D Tidal
Average Screening
Level 2
1-D Seasonal
Steady
2-D Steady with
Overturn
2-D Seasonal Tidal
Average
Level 3
Routing of Daily
Average Hows
Time-Varying
Mass Balance
2-D/ 3-D Inter-
Tidal Quasi-
Dynamic
Level 4
Dynamic Flows
Energy Budget
Circulation
2-D /3-D Intra-
Tidal Flows
4.3 Specific Model Issues for Simulation of Loading and Impacts from Episodic,
Wet-weather Events
Before presenting detailed suggestions for the model selection strategy in Section
4.4, it is desirable to provide a discussion of some of the specific model issues that arise
in attempting to model episodic, wet-weather PS/NPS for TMDLs. Why does TMDL
modeling involving such sources tend to be difficult? Some, among many, causes are
summarized below.
4,3.1. Episodic Nature of Loads
WLA modeling is often conducted with the assumption that loading from sources
is constant or changes infrequently, and that source strength is monitored and known.
This allows use of simulation models using steady-state loading assumptions. Episodic,
wet-weather events can contribute loading to the receiving water in infrequent, short but
intense pulses. The loading function is thus very much dynamic, or non-steady state,
and models which assume steady, or quasi-steady loading rates cannot be expected to
represent the short-time variability of pollutant concentrations. However, this does not
mean that dynamic receiving water models are always required. The fact is that the
quality response of most receiving waters is relatively insensitive to short term variations
4-16
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
in load rate, at least for conventional pollutants. For instance, the response time of lakes
and bays to nutrient loadings is generally on the order of weeks to years, while the
response time of large rivers to OD is on the order of days (Donigian and Huber, 1991).
These relatively long response times mean that steady state receiving water models can
yield appropriate results, if applied correctly, when the pollutant of concern is mixed
into a relatively large area prior to manifestation of the response. On the other hand,
response to acute toxins may require a small time step and careful consideration of the
dynamic timing of the load.
Where steady state receiving water models can be applied, we may need only
average loading rates from wet-weather events. However, the episodic nature of loading
must be taken into account in simulating wetd-weather loading if the receiving water
models demand any more detail than annual average or totals. This usually involves
at a minimum the simulation of the occurrence of wet-weather runoff events from
precipitation. However, in most cases it is sufficient to resolve only to the event total
load scale. The case studies include successful TMDLs performed by linking episodic
loading models to steady state receiving water models. Section 4.4 provides suggestions
on the conditions under which steady state receiving water models with averaged wet-
weather loading can be used, and when, conversely, a dynamic approach to loading
should be used.
4.3.2 Definition of Critical Events
WLAs for constant discharges are often calculated on the basis of a design flow
(e.g., the 7Q10 flow, which is the seven day average low flow with a 10 year return
period). Design flows are chosen to provide a certain degree of protection against water
quality excursions by evaluating effects of the source under the most stressful conditions
which reoccur with a certain specified frequency.
£PA has recommended to the States that water quality criteria statements be made
with an appropriately defined duration (averaging period) and frequency of excursion
(U.S. EPA, 1991). In 40 CFR 122.45(d), EPA requires that all NPDES limits be expressed,
unless impracticable, as both average monthly and maximum daily values. Similarly,
the Guidelines for Deriving National Water Quality Criteria for the Protection of Aquatic
Organisms and Their Uses (U.S. EPA, 1985) recommends a generic form for a water
quality criteria statement.
... aquatic organisms and their vises should not be affected unacceptably
if the four-day average concentration of [pollutant] does not exceed
Vig/L more than once every three years on the average and if the one-hour
average concentration does not exceed yg/L more than once every three
4-17
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
years on the average.
While a frequency-duration format is preferable, a design flow approach is
acceptable for WLAs, because the design flow is translatable into a corresponding
recurrence frequency (U.S. EPA, 1986). Use of design flows thus serves as a surrogate
for a probability of water quality excursions (as defined on a specified duration).
Unfortunately, the direct correspondence between frequency of excursions and
design flows is based on the assumptions that (1) loads are steady, and (2) rate of
loading is not correlated with instream flows. This is clearly not the case for episodic,
wet-weather loads. For example, consider the case of overland NPS BOD loading from
agricultural runoff into a stream. It is evident first that the bulk of the loading occurs
during storm events which yield sufficient precipitation to produce overland flow and
sediment transport to the stream. The loading is not steady, but occurs in intense pulses,
although, in the case of BOD, the effects may be felt for days to weeks afterward.
Further, the same precipitation event which caused the loading to occur is also likely to
increase flow in the stream, and little loading will take place during extended droughts.
Thus, the loading and streamflow are positively correlated with one another. Now
suppose we try to impose a conventional design flow criterion and approximate the total
annual load in one of two ways: If we assume that the load is steady, and divide the
total load among 365 days, the LA resulting from a steady-load design flow may
seriously misrepresent the potential effects that may occur if the bulk of the total
loading actually occurs in a small number of major storm-washoff events. This could
lead to the analysis not being protective of water quality. On the other hand, if we try
to examine the actual episodic loads in combination with the steady-state design flow
criterion, we are likely to be extremely over-protective - because high loads are very
unlikely to occur in combination with low flows.
How then do we determine for episodic sources WLAs/LAs that are likely to
meet WQSs (with a given degree of assurance or probability of exceedance) and will be
protective of designated uses? One alternative would be for the States to formulate and
apply wet weather water quality criteria. This is an alternative which EPA has
encouraged, but which is not currently available in most circumstances. Another
reasonable alternative is to recognize that the design flow approach cannot serve as a
reasonable surrogate for frequency-duration analysis of water quality excursions when
a significant component of the loading to the water body -results from wet-weather,
episodic loads. This implies that one should simulate the actual frequency of WQS
excursions, rather than relying on design flows, when evaluating wet weather-dominated
TMDLs.
Accurate estimation of frequency of excursions of WQSs for waterbodies with
wet-weather dominated loading generally involves continuous simulation over a number
4-18
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
of years of precipitation records, and is a logical way to proceed, when sufficient
resources are available to undertake such an analysis. However, continuous simulation
is not always feasible during scoping or the early phases of a Phased TMDL, due to lack
of data or constraints on available resources to perform the modeling analysis. In such
cases, a quick, but conservative analysis is often needed to provide a preliminary
analysis of the problem, constituting an analogy to the steady-state design condition.
Without simulation of the correlation between wet-weather loads and receiving
water flows, such an approach cannot be expected to yield highly accurate results.
However, for a scoping analysis it is often appropriate to construct a design condition
which consists of (1) a wet-weather event load of a given recurrence frequency,
combined with (2) steady-state or (seasonal) average flows in the receiving waterbody.
In essence, this approach makes a "good guess" at the appropriate design condition,
involving high loads and associated flows (because the highest loads are likely to
coincide with higher than normal flows, use of an average flow is a relatively
conservative approach to determining the appropriate load-flow pair, yet avoids the
obvious inconsistency of combining high, precipitation-driven loads with drought flows).
At the lower, or less complex Levels of analysis described in Section 4.2, we will refer
to this as the combination of a "reasonable maximum" load and a steady-state, average
flow. "Reasonable maximum" can be defined in terms of the desired recurrence interval
of the load; combination with average flows thai results in a somewhat more
conservative analysis of recurrence of WQS excursions (i.e., an inflated MOS),
appropriate to the early phases of analysis.
4.3.3. Stochastic Nature of Loads
In terms of the available assimilative capacity for WLAs, the conceptual equation
for TMDLs (presented in Chapter 1) may be rewritten as
WLAs = UC - Lis - MOS
The analysis is complicated by the fact that the LAs reflect an estimate of a process with
a significant random component. That is, wet-weather loading is a function of storm
intensity, duration, and frequency, which must be described probabilistically. As noted
above, and as EPA has recommended in the WLA guidance, and stated most clearly in
the recent TSD for toxics (U.S. EPA, 1991), criteria statements should be made in a
format with an appropriately defined duration (averaging period) and frequency of
excursion. The fact that criteria and permit conditions take, or are recommended to take
a duration-frequency form requires that modeling for TMDLs should provide similar
information: that is, the modeling and monitoring activities should result in an estimate
of the frequency distribution of receiving water concentrations, rather than just worst-
4-19
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
case and/or average estimates. This is particularly important for episodic, wet-weather
loading, as the concentrations and impacts of these types of sources are sensitive to
variability in both washoff rates and streamflow. Further, in the typical case washoff
rates and receiving water flow are strongly, but not completely correlated with one
another, complicating the analysis.
In the TSD for toxics, EPA (1991) recommends three dynamic receiving water
modeling techniques to yield the equency distribution of receiving water
concentrations: continuous simulation, Monte Carlo simulation, and lognormal
probability modeling.
Continuous Simulation Models combine daily (or other time step) measurements or
synthesized estimates of effluent flows, effluent loads, wet-weather source
concentrations/loads and receiving water flows to calculate receiving water
concentrations. A deterministic model is applied to a continuous time series of these
variables, so that the model predicts the resulting concentrations in chronological order
with the same time sequence as the input variables. This enables a frequency analysis
of concentrations at any given point of interest. The analysis automatically takes into
account the serial correlation that may be present in flows and other parameters, as well
as the cross-correlations between measured variables. This is potentially the most
powerful method available for accurate prediction of the frequency of receiving water
concentrations. However, it does have its disadvantages. Rrst of all, it is very data
intensive. Further, long time series of wet-weather loads will generally not be available,
and these will have to be synthesized from precipitation records. This introduces
uncertainty, and if any other input time series are lacking the uncertainty will likely be
so great that Monte Carlo methods are preferable (see below). EPA (1991) recommends
that if recurrence intervals of 10 or 20 years are to be examined for WQS excursions, at
least 30 years of flow data should be available to provide a sufficient record to estimate
the probability of such rare events.
Monte Carlo Simulation Models combine probabilistic and deterministic analyses.
That is, this approach uses a deterministic water quality model with statistically
described inputs. The model is run repeatedly, at each application drawing a random
realization from the input statistical distributions. .If all the time-varying inputs (such
as flows) and uncertain parameters are described statistically, the result is a simulated
set of receiving water concentrations which reflects the statistical distribution of the
model inputs; however, these concentrations will not follow the temporal sequence of
real data. The Monte Carlo method is potentially very powerful, and requires somewhat
less data than continuous simulation. A particular strength is its ability to provide a
direct assessment of model uncertainty by use of statistical representations of uncertain
parameters.
4-20
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Lognormai Probabilistic Dilution Models have been developed by EPA to provide a
simpler method of frequency analysis. As implemented in EPA's DYNTOX model
(LimnoTech, 1985), the lognormai probabilistic approach takes a simple stream dilution
model, assumes that all the input parameters can be represented by lognormai
distributions, and uses numeric integration to derive the resulting distribution of
receiving water concentrations. While simple to use, the model is limited to application
to rivers and streams, does not include instream fate processes, and cannot correctly
analyze more than a single pollutant lo: ing source. The approach also assumes that
all input parameters, including effluent flow and concentration, can be described by
lognormai distributions. The latter assumption will often not be appropriate for analysis
of impacts of episodic wet-weather loading over a period longer than the average runoff
event, as wet-weather flows can be equal to zero for long periods between events. For
these reasons, lognormai probabilistic modeling will in many cases not be applicable to
calculating TMDLs which involve episodic, wet-weather sources. However, it may be
of use in certain simple cases, such as short-response impacts from a localized wet-
weather source, or situations in which the wet-weather source can be regarded as a
relatively constant background in relation to a dominant point source.
For some cases involving TMDLs with episodic, wet-weather loads, a detailed
prediction of the frequency distribution of concentrations may be desired, but neither a
full.continuous simulation (requiring a long time series of wet-weather load estimates)
nor a full Monte Carlo simulation (which simulates receiving water flows and thus
ignores the effects of serial correlation) may be appropriate. In such cases, hybrid
methods may provide a powerful alternative. For instance, simulation over a continuous
time series of observed receiving water flows could be combined with a Monte Carlo
representation of wet-weather loads. To do this properly, however, the probabilistic
simulation of the wet-weather component would need to take into account the
correlation between wet-weather loading and receiving water flows (since both are
driven by precipitation). For instance, observations could be used to develop a
conditional probabilistic model in which wet-weather loads were simulated conditional
on flow and/or precipitation.
4.3.4. Special Issues for CSO/SW Loading Models
Urban pollutants loaded via CSO/SW. represent ncmpoint washoff processes,
which are however collected in a sewer and discharged to die receiving water as a point
source. Most available dynamic urban loading models combine hydraulic modeling of
the sewer system with modeling of wet-weather contaminant loading to the system.
However, these represent two very different aspects of the process. Because the
hydraulic behavior of sewer systems is well understood, and the physical characteristics
of the system can often be accurately defined, hydraulic calibration can often be
4-21
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
accomplished with relatively high accuracy. On the other hand, the state of the art in
predicting pollutant loading from urban sources is much less satisfactory, and event
mean concentrations may be subject to a high degree of random, or at least poorly
understood, noise. We therefore suspect that it will often be appropriate to undertake
detailed simulation of sewer hydraulic response combined with simpler, spatially and
temporally averaged estimates of pollutant concentrations in order to derive an estimate
of the frequency distribution of pollutant loadings to the receiving water.
4.3.5. Other Loading Mechanisms
In dealing with wet-weather loading, we usually think of wet-weather washoff
processes. However, certain other mechanisms can also be important in the total
delivery of a given pollutant to a water quality-limited waterbody. Pathways to consider
include:
• Secondary generation from sediments: Sediment bound pollutants in
washoff may be deposited episodically, but release to the water column
only gradually. In some cases, the loading may have both a direct
(episodic) and indirect (quasi-steady sediment release) component
• Dissolved constituents in rain itself may be of importance (acid rain).
• For waterbodies with large surface areas, atmospheric deposition of fine
particulates or volatiles may be an important loading mechanism. For
instance, in the Great Lakes, atmospheric deposition is thought to
constitute a major portion of the ongoing loading of PCBs.
• In some climates, wind erosion of sediments may be an important loading
mechanism.
• Dry-weather NPS can sometimes be important, involving loading from
groundwater flux to streams or surface seepage. Diffuse acid mine
drainage is a particularly salient example.
The loading mechanisms addressed in-th» preceding-bullets are not within the
focus of the present guidance. However, their occasional importance should be kept in
mind by the TMDL developer.
This and preceding sections summarize only a few of the key issues of importance
to modeling of the quality of receiving waters impacts by episodic, wet-weather loads.
These and other issues are incorporated into the next section, which attempts to provide
4-22
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
general guidance oil the selection of appropriate models for the calculation of TMDLs
which involve episodic, wet-weather loading.
4.4 Decision Criteria for TMDL Model Identification
We discussed the general criteria for model objectives for TMDLs in Section 4.2,
while Section 4.3 presented in more detail some of the specific technical issues associated
with modeling wet-weather, episodic loading. How can these many factors be combined
into a logical procedure for model identification and selection?
In the following pages we provide an aid to TMDL model identification in the
form of a set of questionnaires and brief decision trees. Working through these steps
will enable the user to come up with a set of formal criteria (e.g., necessity of simulating
time-varying flows) that represent the characteristics of simulation models or scoping
procedures that are appropriate to the waterbody and level of analysis of the TMDL
procedure. The procedure does not recommend specific models, as the realm of
available models is constantly changing and expanding. Rather, it yields a description
of model characteristics which can be matched to the available models summarized in
Chapter 3, or to other models identified by the user.
The decision trees are framed in terms of the Levels of Analysis for TMDLs,
described in Section 4.2 (see Table 4-1), in which it is assumed that the process will begin
at the simplest level that is appropriate, and add complexity as necessary. These were
defined to range from Level 0 analysis (simple scoping) to Level 4 analysis (generally
involving complex dynamic simulation).
The user should note that the decision trees provide only generic suggestions,
which incorporate the authors' opinions and experience regarding common modeling
situations. Specific exceptions to many of these guidelines can no doubt be identified.
Therefore, results of this process should always be reviewed to make sure that they
make sense in terms of site-specific conditions and needs - and modified as necessary.
STEP 0. BASIC CHARACTERIZATION (Questionnaire)
Step 0 consists of forming a catalogue of important data and characteristics to be
utilized in the subsequent decision trees. The data needs are addressed in more detail
in other sections; a recapitulation of the most important items appears below.
0.1. Technical Criteria Data requirements for technical criteria are concerned with the
important physical characteristics of the receiving waterbody, pollutants of concern, and
4-23
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
wet-weather loading sources. These data requirements are discussed in more detail in
Chapter 6. Important points to consider include:
0.1.1 Identify parameters or group of parameters of concern for the present
TMDL. These may include chemical pollutants, sediment, physical habitat
modifications, etc. List constituent(s) under study, and important chemical
and biological characteristics. Does the analysis require consideration of
secondary effects (e.g., W 0 load evaluated in terms of effect on DO
downstream), or do only direct effects (e.g., a numeric criterion for the
discharged constituent) need to be considered?
0.1.2 Catalog known and suspected sources of loading. Are there point sources?
Are there wet-weather, episodic sources of the constituent? What are the
relative magnitudes of the various sources? What is the spatial
relationship of the sources?
0.1.3 If episodic, wet-weather sources are present and of significant magnitude,
what are the associated land use types? What is the pattern of variability
in wet-weather loading? Are nonpoint sources of types which may be
amenable to control through BMPs?
0.1.4 Characterize initial physical boundaries for impact analysis in the receiving
waterbody. Include a descriptive catalog of designated uses, extent of
known impairment, location of any sensitive areas, etc.
0.1.5 Characterize receiving water flow regime. Usually, this question boils
down to whether the receiving water is a river, lake, estuary, etc.
Answering the question is not always as simple as it appears. This issue
is addressed further in Step 4. Basic data requirements are: length scales
of the system, advective velocities, and estimates of dispersion coefficients
and reaction rates. These can be used to form dimensionless numbers and
characteristic mixing times, which index the relative importance of
advection, dispersion, stratification and chemical reactions in receiving
water. Calculation of these numbers is shown in Box 4-1 (following Step
0). Note that more than one receiving water body type may be involved in
the analysis (e.g., a river discharging into an estuary).
0.1.6 Identify chemical/impact parameter characteristics. For chemical
pollutants, an estimate of rates of reaction or decay is often of great
importance (see Bowie et al., 1985 for a basic reference). Is the chemical
preferentially sorbed to sediment or dissolved? Does the chemical
accumulate within the system so that prior years' loadings are important
4-24
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
to understanding present effects? Where physical habitat impairment or
ecosystem response is of concern, how is the wet-weather loading thought
to be expressed in waterbody impairment?
0.2 Regulatory Criteria The regulatory criteria will determine what types of
questions will need to be answered through use of models. These issues are discussed
in more detail in Chapter 1.
0.2.1 What is the format of the relevant WQSs to be addressed by the TMDL?
Do they represent numeric or narrative criteria, or both? If numeric
criteria apply, they may include not-to-be-exceeded values, target
concentrations expressed at critical design flows, or frequency-duration
specifications for long-term averages. Do relevant narrative criteria allow
a one-at-time approach to individual pollutants, or do they imply that a
whole effluent toxicity approach should be followed (e.g., "toxic substances
in toxic amounts")? Is a mixing zone allowed for point sources, and what
criteria apply within and at the boundary, and beyond, of the mixing zone?
0.2.2 If a Phased TMDL is being pursued on the waterbody, what is the stage
of phased TMDL process? Has previous TMDL or WLA analysis been
made of this waterbody-pollutant pair, or of this waterbody and other
pollutants? Was a phased approach undertaken involving BMPs for LAs
which must be reevaluated?
0.2.3 What is the priority ranking for this TMDL?
0.2.4 Do regulations impact the area to be simulated by imposing jurisdictional
boundaries?. The most natural unit for TMDL analysis with wet-weather
loads is the watershed or basin. However, jurisdictional boundaries or
regulatory constraints may preclude treatment of the whole basin within
the TMDL. Describe any regulatory constraints on the boundaries of the
watershed to be simulated in this step.
0.3 User Criteria These include the functional and operational needs of the user,
including limitations of resources available. These issues are discussed further in Section
4.2.
0.3.1 Practical constraints and resources: What is the available level of resources
for this TMDL? What is the time frame for completion?
0.3.2 What models and expertise are available in-house for the analysis? What
is the level of experience available for analysis of similar problems?
4-25
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
0.3.3 Form a preliminary estimate of the acceptable MOS for the TMDL, which
will depend on the level of uncertainty present in the modeling analysis
and monitoring data. The level of uncertainty acceptable in the modeling
analysis should be specified a priori, as noted in the discussion of modeling
DQOs. This is included among the User Criteria because the acceptable
level of uncertainty, or, alternately, the required level of accuracy, will be
an important factor in determining the effort that must be expended to
establish the TMDL. If previous TMDL/WLA analysis has been
undertaken, the level of uncertainty in the earlier work should be
evaluated.
4*26
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Box 4-1. Calculation of Dimensionless Parameters & Characteristic Mixing
Times
The use of dimensionless parameters to characterize receiving water bodies is
largely due to work of Fisher et al. (1979) and Schnoor (1985), and is summarized
in McKeon and Segna (1987).
The relative importance of advection and diffusion is given by the Peclet
number, PE, defined as
PB 9 ulfD
where:
u is the mean velocity (L/T)
Z is the segment length under consideration (L)
D is a dispersion coefficient (L2/T)
This dimensionless parameter is a ratio of the advective transport process to the
dispersive transport process. If the Peclet number is significantly greater than 1.0,
the system is advection-dominated; if it is much less than 1.0/ dispersion
dominates the transport, at least for dissolved conservative substances.
The importance of stratification in an estuary can be evaluated through the
Estuarine Richardson Number, proposed by Fischer (1972):
R. t(Ap/p)' Qh gf
W
where:
R is the Estuarine Richardson Number,
Ap is the difference in density between the river and ocean water (M/L3),
P is the density of the ocean water (M/L3),
g is the acceleration, of gravity (L/T1),
Qf is the freshwater inflow (L3/T),
W is the channel width (L), and
IT, is the root mean square tidal velocity (L/T).
4-27
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Calculation of Characteristic Mixing Times (continued)
observations of the presence of stratification will usually be available. However,
the degree of stratification may change seasonally, and the Estuarine Richardson
Number provides a quick means for evaluating die potential for change, either
seasonally or with altered inflow regimes.
Finally, the residence time of a degrading or reactive contaminant in the water
body has an effect on the importance of the physical processes that must be
modeled. For instance, if degradation occurs much more quickly than advective
transport, it may not be necessary to model the advective transport. These
relationships may be evaluated through the use of characteristic mixing times,
suggested by Eschenroeder (1983):
The advection time is proportional to the principal length scale of the domain
or area of interest, I, divided by the mean velocity, u:
tA - v»
The diffusion time is proportional to the square of the distance scale, W,
divided by the dispersion coefficient, D:
tD - W*PJD
and may be defined for longitudinal dispersion along the main axis of flow (tOL)
or for transverse dispersion normal to the axis of flow (tcrr)*
The importance of chemical reactions can be evaluated via the transformation
time. The transformation time is proportional to the reciprocal of the first-order
rate coefficient, k:
tT - 1 Ik
4-28
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
The following steps provide a series of brief, relatively simple decision trees, in
tabular format. These may be worked through in succession, and are designed to yield
a description of appropriate modeling identification components for a TMDL analysis
of a particular site.
In building a site-specific surface water modeling application we usually start
from the basics (i.e., the flow of water) and add on additional compartments and
complexity as necessary. For the pury se of sorting out a modeling strategy it is,
however, convenient to work in the opposite direction: After an initial determination of
the level of complexity to be pursued, we begin with the top-level components (e.g.,
ecological response) and work backwards to the base level of hydrodynamic simulation.
This is because, in the context of decision for a TMDL, the necessary complexity of all
the component models is driven by the level of complexity required in the response
variable model, which will occupy the top level of the set of models chosen. At each
step we first ask the basic question "Is a model needed?", and, if yes, set forth some
generic decision criteria.
STEP 1. APPROPRIATE LEVEL OF ANALYSIS (Decision Tree)
Runoff
Quantity
TMDL Modeling
Components
Step 1 sets the general framework
for developing an appropriate modeling
strategy for calculation of a TMDL, and,
as such, relates to all the component
models to be employed (see box). The
key to identifying a modeling strategy
which is appropriate not just to the
physical characteristics of the waterbody,
but also to the objectives of the TMDL,
and practical constraints of available resources is to select an appropriate Level of
analysis (see Table 4-1) for the identification of simulation models. The Levels increase
in complexity, and required effort, from Level 0 (simple scoping) to Level 4 (detailed
dynamic simulation). As described in Section 4.2, the Levels may be defined in general
terms as:
Integrated TMDL Model
Level 0
Level 1
Level 2
Simple scoping analysis
Application of simplified modeling representation, typically
involving steady state modeling of waterbody hydrodynamics in
one or two dimensions with annual average or reasonable
maximum wet-weather loads
Elaboration of Level 1 to include factors such as seasonal variability
4-29
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
and a more detailed spatial representation, but still utilizing a quasi-
steady state representation of hydrodynamics
Level 3 Simplified dynamic simulation of receiving water hydrodynamics,
which may involve hydrologic routing of daily or tidally averaged
flows and some form of continuous, event-based representation of
wet-weather loading.
Level 4 Detailed dynamic modeling of receiving waterbody, usually coupled
with continuous simulation of wet-weather loading.
Selection of the appropriate Level of analysis for the current effort begins with the
consideration of any past TMDL analyses:
1- Current state of TMDL analysis for the site and pollutants under examination:
a) New TMDL (no previous analysis for this pollutant on the waterbody); or
preliminary assessment to identify waterbody impairment -> 2
or
b) Previous TMDL work has been undertaken on the waterbody (addressing
the pollutant of present interest). Current effort may involve a
revaluation of WLAs and LAs based on additional monitoring data
obtained under a Phased TMDL approach, or a scheduled review of
permits within a basin -> 3
2. New TMDL for this pollutant/waterbody, or preliminary assessment
a) Established modeling base available for waterbody from prior TMDL or
other work -> 2.1
or
b) No established modeling base -> 2.2
2.1 New TMDL, established modeling base. Evaluate appropriate level of analysis
based on review of previous modeling.
2.2 New TMDL, no established modeling base.
a) High priority TMDL, impairment of special areas of concern, or reported
violations of human health-based WQSs at beaches, drinking water intakes,
and other exposure points -> 2.2.1
or
b) Other -> 2.2.2
2.2.1 High priority new TMDL.
a) Seasonal variability in wet-weather loading significant: Begin analysis at
Level 2
or
4-30
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
b) Impairment of waterbody involves nonlinear interactions among different
pollutants: Begin analysis at Level 2.
or
c) Flow in receiving water body is intermittent: Begin analysis at Level 2
or
d) wet-weather loading (on average) appears to be relatively stable over
seasons, and (b) and (c) do not apply: Begin analysis at Level 1 and
increment if necessary.
or
e) Analysis of excursion of WQSs at human exposure points or special areas
of concern requires analysis of upstream movement of pollutants in an
estuarine system: Begin analysis at Level 3.
2.2.2 Lower priority new TMDL or preliminary assessment.
a) Known impairment of waterbody involves nonlinear interactions among
different contaminants: Begin analysis at Level 2.
or
b) Flow in receiving water body is intermittent: Begin analysis at Level 2
or
c) Tidal estuaries and other systems in which advective and macro-dispersive
forces are both prominent in the transport of pollutant loads. This is
generally characterized by the mixing times tA and tD being of
approximately equal magnitude: Begin analysis at Level 1.
or
d) Other, including preliminary assessments: Begin analysis at Level 0
3. Continued TMDL for a given pollutant and waterbody. Examine results of
previous modeling analyses:
a) Ongoing monitoring shows previous model predictions not within
acceptable range of accuracy -> 3.1
or
b) Prior MOS unacceptably large, or level of uncertainty too high -> 3.1
or
c) Neither (a) nor (b) applies -> 3.2
3.1 Previous TMDL effort needs revision: Increment to next level of analysis, of, if
previously at level 4, reexamine modeling assumptions.
3.2 Previous TMDL level of modeling acceptable. Revisit analysis at same level,
incorporating additional data gained in monitoring.
4-31
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
STEP 2. ECOLOGICAL RESPONSE MODELS
After having established the general
Level of analysis in Step 1, we now return to
a potential top-level component of the
modeling application: ecological response.
Ecological response models are used to p idict
ecological responses to ecosystem degradation
or restoration. They are used to establish
TMDLs/WLAs/LAs to meet habitat criteria
and biocriteria. Many other TMDLs will not
involve ecological response models.
Response
TMDL
Com
Modeling
?onents
Chemical
Transport
Sediment
Transport
Runoff
Quality
Waterbody
Hydrodynamics
Runoff
Quantity
Integrated TMDL Model
1. First, determine if an ecological response model is needed.
a) Physical habitat degradation is of concern in waterbody impairment, or
interpretation of waterbody impairment or restoration requires examination
of responses of biological community, via population dynamics, biomass
production, or biodiversity -> 2
or
b) Evaluation of waterbody impairment or restoration does not require direct
evaluation of biological responses - for instance, the impairment can be
characterized by excursions of numeric criteria for chemical constituents or
by narrative criteria to be evaluated in terms of concentration of the
pollutant, and physical habitat restoration techniques are not under
consideration -> 3
2. Situation requires use of some form of ecological response model (whether
empirical or simulation).
a) Ecological response of interest involves physical habitat modification -> 2.1
or
b) Ecological response of interest involves population dynamics, biomass
production or biodiversity measures -> 2.2
2.1 Habitat modification models. Analysis is often piggybacked on to receiving water
models and/or population dynamic models. Effects usually considered to fall into one
of two broad classes:
a) Alterations in stream morphology (e.g., variations in width and depth,
creation of erosion controls, settling ponds, etc.) -> 2.1.1
or
b) Alterations in stream flow volume or timing, transport dynamics, or
chemical dynamics which alter input to other modeling components (e.g.,
4-32
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
creating reaeration drop structures or planting streambank trees to reduce
solar energy input) -> 2.1.2
2.1.1 Stream morphology models.
• TMDL analysis Levels 0-2: Use empirical methods and guidelines to estimate
impacts, such as those outlined by Leopold et al. (1964), Dunne and Leopold
(1978), Rosgen (1993) and others. Simulation of chemical and sediment transport
may not be needed.
• TMDL analysis Levels 3-4: Consider combination of empirical methods with
simulation of sediment movement in Step 4. Simulation of chemical transport
may not be needed.
2.1.2 Habitat modifications which alter coefficients or inputs to other model
components (such as chemical dynamics in the receiving water, receiving water
hydrodynamics, or biological community models): Evaluate representation in the
appropriate model component in subsequent steps.
2.2 Biological community models. Sophisticated simulation models for biological
communities are generally experimental/research tools with limited predictive ability,
and, at present, little used for TMDL development The primary exception is algal
biomass models, which are frequently applied.
a) Algal biomass models for systems dominated by planktonic algae
(macrophytes and periphytic algae do not play a major role in nutrient
cycling) -> 2.2.1
or
b) Other biomass production and population dynamic problems -> 2.2.2
or
c) Biodiversity or ecological indicator problems -> 2.2.3
2.2.1 Planktonic algal biomass models.
• TMDL analysis Levels 0-1: Employ empirical prediction methods and site
specific correlations to relate biomass to chemical and hydrologic regime. These
will usually require steady-state estimates of flow and nutrient availability.
• TMDL analysis Levels 2-3: Use steady state algal response models to simulate
average response to seasonal average nutrient loading (e.g., algal components of
QUAL2E and similar models).
• TMDL analysis Levels 3-4: Consider applicability of dynamic simulation of algal
response to time varying loads (e.g., algal components of EUTROWASP).
Nutrient concentrations will likely be required on a daily scale.
2.2.2 Other biomass production and population dynamic models.
• TMDL analysis Levels 0-3: Use empirical prediction methods and site specific
4-33
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
correlations. A useful review of models that attempt to predict standing crop of
stream fish is provided by Fausch et al. (1988).
• TMDL analysis Levels 3-4: Consider quasi-steady state (gradually varying)
simulation of population response, given adequate resources and data, in
combination with empirical and observational methods. In many cases the focus
of resources should be on effective monitoring.
2.2.3 Biodiversity and ecological indicator problems. Working simulation models are
generally not available to address this class of problems; instead, a qualitative
assessment is often based on comparison to reference sites. The TMDL developer should
consult EPA's (1990) biological criteria guidance, as well as EPA's series on Use
Attainability Analyses. Depending on the type of pollutant under consideration,
estimates of biological criteria may require average or maximum chemical
concentrations.
3 No ecological response model is needed; proceed to Step 3.
STEP 3. CHEMICAL TRANSPORT MODEL TEMPORAL RESOLUTION
In some cases, the TMDL will be
driven by an ecological response to
chemical concentrations; in others it will
depend on chemical concentrations
themselves. The time step required for
receiving water simulation will depend to
large extent on the temporal resolution
required for chemical or sediment
modeling. We therefore first examine the
receiving water chemical simulation time step and related characteristics. This results
in a decision on whether steady state or dynamic water quality simulation should be
pursued, and, if dynamic, a qualitative indication of the maximum allowable time step.
The spatial resolution for water quality simulation is evaluated in Step 6 along with the
spatial resolution of the hydrodynamic model.
The time step discussed here simply reflects the desired temporal resolution of
the model results, and will typically be considerably larger than the actual simulation
time step used in a dynamic model of water qualitv. This is because most water quality
models use explicit finite-difference solutions, in which case there are stability constraints
on the maximum time step that can be used to ensure model stability (i.e., solution
schemes in which errors introduced in the finite difference approximation tend to damp
Ecological TMDL
Response Com
Modeling
)onents
Chemical Sediment
Runoff
Quality
Waterbody
Hydrodynamics
Runoff
Quantity
Integrated TMDL Model
4-34
-------
The Cadmus Group> Inc.
DRAFT
June 8, 1993
out, rather than amplifying without bound). These stability criteria relate the time step
to the spatial step, flow velocity, dispersion coefficients and decay rate. In these cases,
the size of the time step will be constrained by the spatial resolution required in the
model, among other things. In determining the simulation time step, the user should
consult documentation for the specific model.
1. Is a chemical transport model needed (as distinct from a sediment transport model
alone)?
a) Chemical concentration driven situations: WQS based on dissolved or
suspended concentration; or TMDL based on ecological response expressed
through dissolved or suspended concentration -> 2
or
b) Non-chemical problems involving physical habitat modification, alterations
in flow (without a chemical component), or sediment (as a pollutant itself)
->3
2. Given that a chemical transport model is needed, start by sorting on the Level of
the TMDL analysis.
a) Concentration-driven analysis at Level 0,1, or 2; qi ecological response-
driven analysis requiring averaged concentrations (as opposed to a
continuous simulation of concentrations at points in space and time) -> 2.1
or
b) Concentration driven analysis at Level 3 or 4; gr ecological response-driven
analysis which requires a description of concentrations at specific points in
space and time ->> 2.2
2.1. Analysis at Levels 0 through 2. For the simpler analyses it is often, although not
always, possible to use steady-state water quality approximations. The analysis can be
based on average concentrations (for long response times) or reasonable maximum
concentrations (for short response times). Note, however, that use of the steady state
approximation generally means that an accurate estimate of the frequency of excursion
of WQSs cannot be obtained; therefore the MOS must be increased to account for this
uncertainty. The alternative, dynamic (time-varying) simulation of water quality (Levels
3-4) can provide more accurate results, but at the cost of considerably more modeling
effort. The error introduced by the steady-state assumption, and possible means of
compensating for it, differ widely with the type of waterbody under consideration.
Therefore, next identify water body type (hydrodynamics and transport properties) as
dominantly 1-D advective (e.g, river), dispersive (e.g., lake), or 2,3-D advective-dispersive
(e.g., estuary).
a) Typical river hydrodynamics: dominantly advective system with PE» 1
or region of interest -> 2.1.1
or
4-35
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
b) Typical lake/reservoir hydrodynamics: dominantly dispersive system with
PE« 1, and tDL and tar « tA -> 2.1.2
or
c) Situations neither clearly (a) nor (c), including estuaries, complex and wide
rivers, etc. -> 2.1.3
2.1.1 Level 0-2 analysis for rivers (dominantly advective systems). For advective
systems, an assumption of steady k ding will often provide a relatively good
representation of the maximum concentration response to a single episodic or slug input,
whether analysis is required for near field concentrations, far-field concentrations, or
secondary effects. This is because the pollutant is advected away from the source. In
the far-field, assumption of a steady source is then equivalent to examination of the
behavior of a discrete slug under the assumption that the concentration is not reduced
by longitudinal dispersion. The simplifying analysis thus has a built in conservative
assumption. For near-field analysis, the steady state assumption is merely a dilution
calculation (with reactions considered if appropriate), and the steady representation of
a slug input is again conservative. When multiple episodic sources are considered, the
steady state approximation is equivalent to assuming that the peak concentrations from
all sources coincide, which may be extremely conservative. General suggestions are:
• Levels 0-1: Use steady state water quality simulation.
* Level 2: Use seasonally steady water quality simulation, or superpose analytical
solutions to the advection-dispersion equation for slug or episodic loads on a
steady state hydrodynamic model.
2.1.2 Level 0-2 analysis for lakes and impoundments (dominantly dispersive systems).
Short term variability in water quality in dispersive systems subjected to episodic loads
is not as readily described by steady state analysis as advective systems. On the other
hand, the response time of lakes to many pollutant inputs tends to be long, in which
case analysis of average rates of wet-weather loading is often sufficient. It is important
to examine the relative length of the response time of the waterbody compared to the
average duration of a loading event. Are we looking at a short-term (acute) or long-term
(chronic) response? For some pollutants, both pathways may need to be followed (e.gl,
when concentration limits for both shorter and longer averaging periods are included
in a WQS). The criteria can also be applied to analysis of secondary or indirect effects,
such as effects of BOD loading on a DO criterion. The user should, however, base the
analysis on the response time of the parameter of interest (in this example, DO).
a) Numeric criterion with averaging period less than, or of the same order of
magnitude as the average interevent time between loading events; 0£
evaluation of short-term (acute), non cumulative direct effects of pollutant -
> 2.1.2.1
and/or
b) Numeric criterion with averaging period greater than the average
4-36
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
interevent time between loading events; or evaluation of long-term, or
cumulative effects where the response time of the waterbody is greater
than the average inter-event loading time -> 2.1.2.2
2.1.2.1 Level 0-2 analysis for lakes, "short" response, where spatial concentration
gradients must be considered to obtain a reasonable estimate of excursions of WQSs, and
thus full lateral mixing of episodic loads is unreasonable to assume. This includes both
near-field dilution and subsequent part i mixing into the epilimnion. For dominantly
dispersive systems, a steady loading approximation is not useful for episodic loads.
Instead, screening approximations can be based on analytical solutions to the 2-D
diffusion equation. Use of analytical solutions avoids the problem of specifying a
simulation time step. For lakes, the episodic wet-weather inputs will generally be
represented by a set of tributary discharges, and thus can be represented as intermittent
point sources.
• Level 0: Use time-varying diffusion equation quality solution with worst case
or reasonable maximum event loading. (Superpose solutions from different loads.)
• Levels 1-2: Use time-varying diffusion equation quality solution with observed
or simulated loads.
2.1.2.2 Level 0-2 analysis for lakes, "long" response. The long response time means
that a steady state analysis, based on average loading, can be used for die initial Levels
of analysis. (Problems in spatial resolution for this approach are addressed in Step 5).
The analysis can be based on a steady-state water quality simulation with average loads
per year or season, or maximum average load over the response period, as appropriate
to the pollutant under study. Do we need to consider accumulation of the pollutant in
the system? This is often the case for low solubility, sediment bound toxics which may
release slowly to the water column and result in a very slow response to changes in
loading.
a) Role of accumulation is not of major significance; responses to change in
loading should be fully expressed within a year -> 2.1.2.2.1
or
b) Role of accumulation needs to be considered -> 2.1.2.2.2
2.1.2.2.1 Level 0-2, "long" response without significant accumulative effects. Use
modeling based on annual or seasonal average loading. As above, a steady state
solution can readily incorporate first-order reactions, given steady loading and advective
and dispersive fluxes in the receiving water body. Attainment of numerical
concentration standards may be estimated by steady-state dilution of average loading,
with sufficient MOS to account for year to year variability. For secondary effects,
particularly eutrophication problems, the following water quality temporal
representations can be used (see Mandni et al., 1983):
• Level 0: Preliminary scoping can often be accomplished by using empirical or
4-37
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
regression approaches based on average loading. This is particularly well
developed for analysis of lake trophic response to nutrient loads.
• Level 1-2: Steady state water quality simulations are generally appropriate,
based on average loading rates.
2.1.2.2.2 Level 0-2, "long" response with significant accumulative effects. As in
2.1.2.2.1, steady state water quality simulations are usually appropriate; however, bed
interactions may result in changes in loading taking a long time to reach steady state.
Calculation of time to steady state is presented in Hydroqual (1986).
2.1.3 Level 1-2 analysis for estuaries and other advective-dispersive systems. (Level 0
analysis is not appropriate for this type of complex system.)
a) WQSs or responses require analysis of near field (mixing zone)
concentrations or response for at least some loading sources -> 2.1.3.1
and/or
b) Far-field analysis required -> 2.1.3.2
2.1.3.1 Level 1-2 estuarine analysis including near field response. Evaluate
dilution of discharge using steady state water quality mixing zone models. If far-field
analysis is also required, complete simplified far-field analysis (2.1.3.2) and evaluate
potential impact of far-field concentrations on local mixing zone concentrations.
2.1.3.2 Level 1-2 estuarine analysis for far-field responses. Steady state or quasi-
dynamic simulations of water quality are appropriate at these Levels, utilizing loading
estimates which are averaged at a time frame longer than the tidal cycle (e.g., daily to
yearly). Appropriate water quality models are:
• Level 1: Apply steady state water quality analysis driven by annual or seasonal
average loading to obtain an approximate prediction of annual or seasonal mean
conditions.
• Level 2: Water quality simulation may be steady-state or tidally averaged
(quasi-dynamic). Quasi-dynamic models should be able to predict water quality
variations on the order of days to months.
2.2. Analysis at Level 3-4. These levels will generally involve time-varying (dynamic)
simulation of water quality and pollutant routing in the receiving waterbody. The
intention is to predict the probability distribution of receiving water concentrations,
rather than just a worst-case or average concentration. EPA (1991) has recommended
three dynamic modeling techniques for establishing WLAs: continuous simulation,
Monte Carlo simulation, and lognorxnal probability modeling. However, lognormal
probability modeling is generally not applicable to TMDLs involving episodic, wet-
weather loads, as described in Section 4.3.3.
a) Analysis at Level 3 -> 2.2.1
4-38
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
or
b) Analysis at Level 4 -> 2.2.2
2.2.1 Analysis at Level 3: Quasi-dynamic water quality simulation is usually applicable.
For rivers, this will usually mean modeling quality on a day to day basis, while for lakes
a time-varying mass balance approach can be used with a maximum time step of days
to weeks. For estuaries, a tidal period average time step is usually appropriate. The
preferred type of dynamic or quasi-dynamic analysis can be determined as follows (see
EPA, 1991):
a) Relatively complete time series of receiving water flow, and daily rainfall
and evaporation (to estimate runoff), or direct measurements of runoff,
representative of current conditions are available for a period
approximately twice as long or longer than the return period that must be
predicted; and sufficient data are judged available to obtain an adequate
calibration of all model parameters: Continuous simulation over the
historic time series is preferred as the most powerful technique to estimate
frequency of WQS excursions.
or
b) Time series records are not sufficiently long; qt significant alterations in the
system have taken place so that historic receiving water flows are not
representative of current conditions; fir role of parameter uncertainty must
be examined: Monte Carlo simulation may be preferable, or a combined
Monte Carlo-continuous approach (see Section 43.3). Note that the time
step for full Monte Carlo simulation input should be defined as the
averaging time in the relevant criterion under investigation.
2.2.2 Analysis at Level 4: Use fully dynamic water quality simulation. In terms of the
requirements of the TMDL analysis, the maximum receiving water chemical time step
should be the smaller of the averaging time specified in the criteria statement and the
typical duration of a loading event. For cases in which the specified averaging time is
greater than the typical event duration, the maximum time step is the event scale. The
actual simulation time step will usually be smaller than this maximum value, due to
model stability criteria, desired spatial resolution, etc. Determine the appropriate
dynamic simulation method (continuous or Monte Carlo) as in 2.2.1.
3. No chemical transport model required.
4*39
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
STEP 4.
SEDIMENT TRANSPORT MODEL
This step evaluates the need for
and general type of sediment transport
modeling to be employed in the receiving
waterbody. Sediment transport will be of
concern when it is a significant factor i:
transport of the pollutant of interest, or
when sediment itself is of concern, either
as a pollutant or a driving factor in
morphological change.
Ecological
Response
TMDL Modeling
Components
Chemical
Transport
Sediment
Transport
Runoff
Quality
Waterbody
Hydrodynamics
Runoff
Quantity
Integrated TMDL Model
1.
Is a sediment transport component needed?
a) Sediment itself is pollutant of concern -> 2
or
b)
or
c)
or
d)
or
e)
Stream morphology is to be modeled (assumed applicable only to
streams/rivers)
-> 3
Pollutant of concern tends to be sorbed to sediment or particulate matter.
This includes hydrophobic organic compounds and most metals -> 4
Level 3-4 simulation of IX) problems where sediment oxygen demand is
significant, or Level 3-4 simulation of secondary Water quality effects
significantly influenced by cycling of sediment, organic carbon or
particulate organic matter -> 4
Pollutant of concern is dominantly dissolved and transport in receiving
waterbody is not controlled by movement of sediment or organic matter -
>5
2. Sediment TMDL: Sediment component is needed. Return to Step 3 and analyze
the time step based on sediment as a "chemical" parameter. For the purposes of the
analysis of modeling strategy, treat sedimentation as a first order "decay" process.
3. Stream morphology TMDL, which may be based on empirical analyses or
simulation of sediment transport.
a) Analysis at Levels 0-2 -> 3.1
b) Analysis at Levels 3-4 -> 3.2
4-40
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
3.1 Stream morphology TMDL at Levels 0-2. If changes in morphology are evaluated
via empirical relationships no sediment model component will be needed.
3.2 Stream morphology TMDL at Levels 3-4. Sediment transport modeling may be
required to support any stream morphology modeling determined in Step 2. Usually,
only averaged measures of transport will be required. However, as bulk sediment
movement is largely driven by infrequent flood events, a detailed analysis of
hydrodynamics may be needed.
4. Strongly-sorbed pollutant simulation, or other situation in which movement of
sediment and/or organic carbon will be needed to support the chemical simulation.
a) Analysis of organic compounds which preferentially sorb to organic carbon
or organic matter; or dynamic analysis of sediment oxygen demand -> 4.1
or
b) Analysis of metals or other ionic compounds whose movement is
associated with the movement of sediments -> 4.2
4.1 Simulation of strongly-sorbed organic pollutants. Accurate representation will
need to consider organic carbon as a state variable. This may further need to be divided
into particulate and dissolved organic carbon, the first of which is driven by fine
sediment-transport and exchange with the benthos. Algal organic carbon production
may also need to be simulated for a full representation. Such complex models will
usually not be feasible unless substantial resources are available; they are thus
inappropriate for the simpler Levels of analysis.
• Levels 0-2: Dynamic simulation is not feasible at these levels; therefore detailed
models of sediment and organic carbon transport are not needed. Sorbed
transport may be approximated by consideration of sediment rating curves,
literature partition coefficients, etc. - with the recognition that significant errors
in prediction may result.
• Levels 3-4: The more advanced levels of analysis will usually involve dynamic
or quasi-dynamic simulation of water quality. Where sorbents play an important
role they should also be simulated, generally at the same time scale as the
hydrodynamic model. Chemical reactions will also generally need to be
considered.
4.2 Analysis of contaminants, such as many metals, .associated with inorganic
sediment component. Redox reactions, metal speciation, precipitation and dissolution
may all need to be considered for a detailed understanding of the problem. Sediment
transport can be simulated with a degree of sophistication similar to that described
under 4.1.
5. No sediment transport component is needed for the TMDL.
4-41
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
STEP 5. HYDRAULIC MODEL TIME STEP/SOLUTION METHOD
This step addresses the time scale
of the receiving water hydrodynamic
component of the simulation, including
whether a steady state or dynamic
representation should be used. The actual
simulation time step used in a model
application will also involve consideration
of chemical dynamics and regulatory
criteria.
1. Level of present analysis:
a) Level 0 scoping analysis, not including tidal estuaries -> 2
or
b) Subsequent level -> 3
2. Level 0 scoping analysis. Begin scoping with steady state mass balance or
empirical approaches (no detailed simulation of flow). For instance, lake eutrophication
problems might be scoped with regression methods relating trophic status to nutrient
loading, while concentration of toxics could be estimated by dilution calculations.
3. Level 1 or greater TMDL analysis. Identify water body hydrodynamics and
transport properties as dominantly 1-D advective (e.g, river), dispersive (e.g., lake), or
2,3-D advective-dispersive (e.g., estuary).
a) Typical river hydrodynamics: dominantly advective system with PE » 1
or region of interest ->3.1
or
b) Typical lake/reservoir hydrodynamics: dominantly dispersive system with
Pg 1, and tpL and tpj « t^ -> 3.2
or
c) Situations neither clearly (a) nor (c), including estuaries, complex and wide
rivers, etc. -> 3.3
3.1. Rivers and other dominantly 1-D advective systems
Level 1,2 analysis ->3.1.1
or
Level 3,4 analysis -> 3.1.2
3.1.1 Level 1,2 analysis for rivers. Flow routing will usually not be necessary.
Ecological
Response
TMDL Modeling
Components
Chemical
Transport
Sediment
Transport
Runoff
Quality
^1*
- ¦ -
$mi£8
Runoff
Quantity
Integrated TMDL Model
4-42
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
• Level 1: Use steady description of flow; select appropriate worst-case design
condition
• Level 2: Use steady description of hydraulics with separate description for each
season.
3.1.2 Level 3,4 analysis for rivers. Unsteady (time-dependent) flow effects should be
considered.
• Level 3: Set maximum time step to smaller of maximum chemical time step
(from Step 3), WQS averaging time, or 1 day. Route daily average flows using
hydrologic or hydraulic routing; interpolate from daily flows if needed.
• Level 4: Simulate with hydraulic (momentum) routing. Appropriate maximum
time step is likely to be determined by desired spatial resolution of output and
model solution stability criteria relating time step to spatial increment
3.2 Dominantiy dispersive systems, primarily lakes. In these systems, the appropriate
hydrodynamic representation at a given Level of analysis is generally the same whether
nutrients, toxics, near-field, or far-field problems are under consideration.
a) Level 1,2 analysis -> 3.2.1
or
b) Level 2,3 analysis -> 3.22
3.2.1 Level 1,2 analysis for dominantiy dispersive systems. At these levels of analysis,
lakes can generally be represented by a steady state mass balance, with constant inflow
and outflow to represent advective sources and sinks (Mancini et al., 1983). Within lake
flows are generally not represented. However, at Level 2 it may be useful to consider
seasonal variability in volume and rates of inflow and outflow. For short-response
problems, the steady internal hydrodynamics may be combined with assumptions of
reasonable maximum loading in inflows, or frequency distribution of such inflows.
3.2.2 Level 3,4 analysis for dominantiy dispersive systems. At these Levels of analysis,
time variability should be taken into account. However, because of the relatively slow
response time of lakes to many pollutants, the time step may still be large compared to
that required for simulation of advective systems. For any changes in loading, time to
steady state may be quite long for many lakes, particularly where bed sediment
interactions are important. In the WLA guidance, Hydroqual (1986) provides methods
for calculation of time to steady state for simulation modeling of lakes.
• Level 3: This level corresponds to the Time-Varying Mass Balance models for
lakes described by Mancini et al. (1983) and Hydroqual (1986). The lake may be
represented as completely mixed, or as a set of mixed segments. Time variability
in inflows and outflows (from the lake or between segments) is generally
represented on a scale of days to weeks. This level of hydrodynamic
representation can often be used to support at least approximate water quality
4-43
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
simulations at a somewhat smaller time step, if needed.
• Level 4: This level comprises dynamic models of circulation within lakes. The
hydrodynamic simulation will generally require use of hourly or sub-hourly
maximum time step sufficient to model insolation and wind energy induced
circulation patterns for continuous simulation.
3.3. Advective-dispersive systems, including most estuaries, some wide or
hydraulically-complicated rivers, short- sidence time reservoirs, etc.
a) Systems with tidal boundary (estuaries) -> 3.3.1
or
b) Other advective-dispersive systems -> 3.3.2
3.3.1 Estuaries
a) Level 1,2 analysis -> 3.3.1.1
or
b) Level 3,4 analysis -> 3.3.1.2
3.3.1.1 Level 1,2 analysis for estuaries. Investigate using steady state
approximations with tidal mixing approximated as effective dispersion.
• Level 1: Use steady description of flows based on observations and mass
balance; select appropriate worst-case design conditions.
• Level 2: Use seasonally steady description of flows, perhaps based on
hydrologic routing and allowing for seasonal changes in stratification where
appropriate.
3.3.1.2 Level 3,4 analysis for estuaries.
• Level 3: Use daily inflows and tidally averaged internal flows to build a daily
average advective-dispersive model. Simulation time step will be determined by
model solution stability and desired spatial and temporal resolution of results.
• Level 4: Use hourly or sub-hourly maximum time step to model intra-tidal
variability. Much shorter internal time steps (e.g., seconds) will be required many
models to maintain stability in solutions.
3.3.2 Other (non-tidal) advective-dispersive systems. Exact determination of hydraulic
time step will depend on site-specific conditions. General suggestions are:
• Level 1: Use annual average steady state approximation, with consideration of
worst case design flows, where feasible.
• Level 2: Continue steady state analysis, but consider seasonal variability.
• Level 3: Drive time-varying model with daily average flows; set maximum time
step to smaller of chemical time step, WQS averaging time, or 1 day.
• Level 4: Set time step sufficiently small to allow dynamic simulation of
hydraulics.
it-it*
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
STEP 6. RECEIVING WATER MODEL SPATIAL REPRESENTATION
This step is concerned with
determining the type and dimensionality
of the receiving water simulation models -
which include both hydrodynamic anc
water quality (chemical and/or sediment)
components. Length of the spatial step
will be influenced by prediction needs as
well as model solution stability in
dynamic simulation models.
1. Is a far-field receiving water model needed?
a) TMDL considers far-field effects outside of the mixing zone or immediate
area of wet-weather discharge; near-field effects may also be considered ->
2
or
b) TMDL considers near-field effects only, where only concentrations close to
point of discharge into the receiving waterbody, or in a defined mixing
zone are of concern. Situations in which this may apply include WQSs
which address only mixing zone concentrations and narrative criteria for
reactive substances which decay with sufficient rapidity such that
essentially no transport occurs in the waterbody. The latter condition can
be expressed in terms of characteristic mixing times as tT « tA and tT «
tD -> 3
2. Far-field receiving water models. State of TMDL analysis is at:
a) Level 0, not including tidal estuaries -> 2.1
or
b) Estuaries, and Level 1 or greater for other water bodies -> 2.2
2.1. Level 0 analysis. Use screening mass-balance approach in which the waterbody
is treated as a zero-dimensional stirred reactor (for lakes/dispersive systems) or as a
string of linked zero-dimensional segments (for rivers/advective systems).
2.2. Identify water body hydrodynamics and transport properties as dominantly 1-D
advective (e.g, river), dispersive (e.g., lake), or 2,3-D advective-dispersive (e.g., estuary).
This may be done using the dimensionless numbers and characteristic mixing times
developed in Step 0.
a) Dominantly advective systems, typically rivers. In general, PE » 1, tA «
Runoff
Quality
Runoff
Quantity
Integrated TMDL Model
Ecological
Response
TMDL Modeling
Components
4-45
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
tjyjv and tA « tDL for the region of interest -> 2.2.1
or
b) Dominantly dispersive systems, typically lakes. Generally characterized by
Pg « I, tp « tA 2.2.2
or
c) Complex systems in which both advection and dispersion are of the same
general order of magnitude in determining contaminant distribution.
Typically tidal estuaries -> 2.2.3
2.2.1. Dominantly 1-D advective system
a) No significant stratification, vertically well-mixed -> 2.2.1.1
or
b) Stratification is significant (rare in true rivers; may apply to short-residence
reservoirs) -> 2.2.1.2
2.2.1.1. Unstratified, 1-D advective system
a) Perennial flow throughout region of impact -> 2.2.1.1.1
or
b) Intermittently flowing streams -> 2.2.1.1.2
2.2.1.1.1.1-D hydrodynamic and water quality simulation appropriate beyond zone of
lateral mixing. If WQS evaluation is required within mixing zone, or in regions of
complex flow junctions, 2-D analysis may be required.
• Levels 1-2: Use input description of flows or steady state 1-D hydrologic
simulation, coupled with 1-D simulation of quality. Consider uncertainty
introduced by assumption of complete lateral mixing.
• Level 3: 1-D simulation of dynamic hydrology with 1-D water quality
simulation Qaterally averaged).
• Level 4:1-D dynamic hydraulic simulation with 2-D water quality simulation.
2.2.1.1.2. Intermittent flow streams. Same analyses as 2.2.1.1.1 generally appropriate for
periods of flow; however, special techniques may be required to model dryout phase,
including ground water interactions.
2.2.1.2. Stratified systems with dominantly advective flow. Hydrodynamics are
essentially 2-D (vertical) and water quality stthuiation may Involve transport between
upper and lower levels; 1-D analysis of the layer receiving discharge is likely sufficient
for the scoping levels of analysis (Level 0-1). Beyond Level 1, use of a 1-D water quality
analysis should be explicitly justified.
• Level 1: Use input description of flows or 1-D steady routing, with 1-D
simulation of water quality in upper layer. Transport between layers should be
characterized as net source, net sink, or no net interaction.
4-46
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
• Level 2: Input description of flows or use quasi-2-D steady routing (2 layers),
with quasi-2-D water quality simulation (1-D representation of each layer with
estimated fluxes between layers).
• Level 3: Quasi-2-D dynamic simulation of flow and quality.
• Level 4: Quasi-2-D dynamic simulation of flow, 3-D simulation of water quality.
2.2.2. Dominantly dispersive systems, typically lakes, which may show thermal or
density stratification.
a) Stratification generally maintained for all or most of the summer -> 2.2.2.1
or
b) Unstratified, or weakly stratified and subject to frequent overturns ->
2.2.2.2
2.2.2.1 Stratified, dominantly dispersive system
a) Laterally well mixed, in terms of pollutant and averaging time of interest.
Lateral concentration gradients are either not significant or dissipate in less
time than appropriate for averaging time for numeric standards or
waterbody response time for narrative criteria -> 2.2.2.1.1
or
b) Not laterally well mixed; (a) does not apply -> 2.2J2.1.2
2.2.2.1.1 Laterally well mixed, stratified, dominantly dispersive system. 1-D vertical
analysis is sufficient to describe average distribution of pollutant
• Levels 1-2: Input or simulate steady 1-D vertical description of flow, 1-D vertical
simulation of water quality, is generally appropriate.
• Level 3: Use 1-D (vertical) dynamic simulation of hydrodynamics; 1-D or 2-D
(longitudinal-vertical) simulation of water quality.
• Level 4: 2-D or 3-D simulation of hydrodynamics and water quality may be
necessary to reproduce the temporal effects of vertical mixing at the most detailed
level of analysis.
2.2.2.1.2 Not laterally well mixed in terms of averaging time or response of interest,
stratified dominantly dispersive system. The distribution of pollutants and
hydrodynamics of the system are inherently three-dimensional, but simpler
approximations may be useful.
• Levels 1-2: Input or simulation steady 1-D vertical description of flow, use 1-D
vertical simulation of quality on most-impacted segments of the waterbody. This
incorporates conservative assumptions by neglecting diffusion out of most-
impacted segments to regions of lower concentration.
• Level 3: Improve analysis by using 2-D (vertical) representation of water quality
of most-impacted segments with specified concentration or flux boundary with
remainder of waterbody.
4-47
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
• Level 4: 2-D or 3-D simulation of hydrodynamics with 3-D simulation of water
quality in whole waterbody or portion of waterbody.
2.2.2.2 Unstratified dominantly dispersive system; pollutant discharges are mixed
vertically within a time frame that is short relative to the needed analysis.
a) System is laterally well mixed, in terms of pollutant and averaging time of
interest -> 2.2.2.2.1
b) System cannot be consider i laterally well mixed -> 2.2.2.2.2
2.2.2.2.1 Laterally well mixed, unstratified, dispersive systems. Pollutants are well-
mixed throughout the volume of the waterbody within the response time scale of
interest. A zero-dimensional (completely stirred reactor) simulation will likely be
appropriate for most applications.
2.2.2.2.2 Not laterally well mixed, unstratified, dispersive systems. Because the
system is vertically mixed, the simulation of water quality is naturally 2-D (plan view).
• Levels 1-2: Input steady zero-dimensional description of flow (input-output).
Consider use of zero-dimensional representation of most-impacted segments of
the waterbody. This incorporates conservative assumptions by neglecting
diffusion out of most-impacted segments to regions of lower concentration.
• Level 3: Improve analysis by using 2-D (plan view) water quality representation
of most-impacted segments with specified concentration or flux boundary with
remainder of waterbody.
• Level 4: 2-D simulation of hydrodynamics; detailed 2-D or 3-D simulation of
waterbody or portion of waterbody.
2.2.3. Advective-dispersive systems, including most estuaries. For estuaries, the concept
of Levels of Analysis 0-4 presented here is similar to Levels 1 through IV of estuarine
simulation models described by Ambrose et al. (1990).
a) Laterally well mixed in direction transverse to main axis of flow, in terms
of pollutant and averaging time of interest. Concentration gradients are
not significant across axis of flow. This will generally mean that tpr < tA -
> 2.2.3.1
or
b) Not laterally well mixed -> 2.2.3.2
2.2.3.1 Laterally well mixed, advective-dispersive system.
a) Strongly stratified system. For estuaries, Fisdier et al. (1979) suggest that
the transition from vertically well-mixed to stratified estuaries generally
occurs in the range of 0.08 < R < 0.8, where R is the estuarine Richardson
number (see Step 0). For R > 0.8, the system can be treated as strongly
stratified. Within the transition range, the determination should be based
4-48
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
on observation -> 2.2.3.1.1
or
b) Vertically well-mixed system -> 2.2.3.1.2
2.2.3.1.1 Laterally well mixed, stratified, advective-dispersive system. The
hydrodynamics of the system are essentially 3-D, while pollutant transport can be
described as 2-D (vertical); however, simpler analysis may be sufficient for many
problems.
• Levels 1-2: 1-D analysis of the upper layer is usually sufficient, although the
effects of ignoring interactions with lower layer should be considered. Note that
the lower layer, adjacent to the sediments, may be a net source of certain
pollutants (such as BOD).
• Level 3: 2-D (vertical) simulation of circulation could be used to improve
resolution.
• Level 4: Detailed representation of transport processes likely needs a 3-D
hydrodynamic simulation.
2.2.3.1.2 Laterally well mixed, unstratified, advective-dispersive system.
Hydrodynamics can be described in two dimensions, while contaminant distribution
varies significantly in one direction only.
• Levels 1-3:1-D simulation of hydrodynamics suggested as adequate for analysis
purposes, except insofar as the laterally well mixed assumption is not exactly true.
• Level 4: 2-D (vertical) or 3-D simulation of hydrodynamics can capture more
of the full spatial and temporal variability of transport processes.
2.2.3.2 Not laterally well mixed advective- dispersive system.
a) Strongly stratified system. For estuaries, Fischer et al. (1979) suggest that
the transition from vertically well-mixed to stratified estuaries generally
occurs in the range of 0.08 < R < 0.8, where R is the estuarine Richardson
number (see Step 0). For R > 0.8, the system can be treated as strongly
stratified. Within the transition range, the determination should be based
on observation -> 2.2.3.2.1
or
b) Vertically well-mixed system -> 2.2.3.2.2
2.2.3.2.1 Stratified, not laterally well mixed, advectiverdispersive system. Both
hydrodynamics and pollutant transport are inherently three-dimensional.
• Levels 1-2: Obtain approximate results with a 1-D analysis (hydraulic and
transport) of the receiving layer. The fact that lateral variability and potential role
of input from lower layer is ignored will result in the need to develop an estimate
of potential errors in inherent in the approximation and either assign a large MOS
or move to a more sophisticated analysis.
4-49
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
• Levels 3-4: A two-dimensional approximation or a full three-dimensional
analysis of the estuary is needed. See EPA's estuarine WLA guidance (Ambrose
et al., 1990) for further details.
2.2.3.2.2 Unstratified, not laterally well mixed advective-dispersive system. 2-D
analysis generally required.
3. Near-Field modeling only required for TMDL. This implies that the important
sources of pollutants resulting in water quality limited waters are separable, and do not
interact with one another in any significant way. The TMDL thus reduces to a set of
wasteload allocation problems.
STEP 7. WET-WEATHER LOADING
Choice of wet-weather loading
models will first reflect the physical
characteristics of the system, such as land
use and pollutant types. For instance,
urban and non-urban wet-weather
loading problems are typically simulated
with different models. However, certain
general considerations apply to all types
of wet-weather loading models.
In the context of TMDL development, the role of wet-weather source modeling
and monitoring is to provide input to a receiving water quality analysis of waterbody
impairment - whether in the form of an estimate of the total load delivered as a driver
for a receiving water quality model or analysis. We thus need "end of the pipe" type of
measurements. From this narrow viewpoint, we are not necessarily interested in the
details of land surface processes, except as those details help improve estimates of
loading. In addition, the spatial and temporal resolution of our loading predictions does
not need to be any finer than is required to drive the receiving water analysis.
Given these caveats, it is clear that wet-weather load modeling for TMDLs will
often be quite different from state-of-the-art descriptive "simulation models. The
interplay between monitoring and modeling is particularly important here: in many
cases, the best course of action may involve site specific determination of pollutant
concentrations via monitoring coupled with an event-total simulation of runoff volume,
thereby yielding an estimate of the time series of loads. Therefore, we first need to
inquire whether any wet-weather load model is needed, and, if so, whether quantity and
quality, or just quantity should be simulated.
MODELS - GENERAL
Ecological TMDL
Response Com
Modeling
jonents
Chemical Sediment
Transport Transport
Waterbody
Hydrodynamics
Integrated TMDL Model
4-50
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
1. Are any wet-weather loading models needed? In some cases, no modeling
analysis at all may be needed for the wet-weather sources. This will generally be when
characterization through monitoring is adequate to answer the questions at hand.
a) Wet-weather loads are relatively constant, well-characterized by
monitoring, and not amenable to or not considered for control in this phase
-> No wet-weather load modeling; rely on monitoring data.
or
b) Scoping analysis at Level 0, with adequate monitoring data base for
preliminary analysis -> No wet-weather load modeling for the preliminary
analysis.
or
c) Other situations, requiring wet-weather load modeling -> 2
2. Wet-weather load modeling needed. It is assumed that this will require some
degree of quantity (runoff) modeling. The quality component may or may not be
addressed by modeling.
a) Receiving water model requires loading on annual, average or event-mean
basis (not intra-event); the source is susceptible to accurate monitoring
(e.g., CSOs) and event-mean concentrations are well characterized by
monitoring; and extrapolation beyond the monitoring record is not
required at this stage -> User should consider representing wet-weather
loading via analysis/simulation of flows (quantity) coupled with estimates
of event mean concentration derived from monitoring. If so, only quantity
modeling will be needed for wet-weather loads at this level, complete Step
7A and 7B
or
b) Other situations, in which modeling of wet-weather loads is needed ->
Wet-weather quantity and quality modeling needed, complete Steps 7A,
7B, 8A and 8B
4-51
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
STEP 7 A WET-WEATHER QUANTITY MODELING - TIME STEP
The temporal resolution for
modeling of wet-weather quantity (flows)
will be largely determined by the needs
of the receiving water analysis.
1. Establish the appropriate level of
temporal detail for hydrologic modeling
of wet-weather load contributions.
a) Chemical time step for
receiving water established in Step 3 is steady, or annual or seasonal
average -> 2
or
b) Chemical time step for receiving water is continuous, event or sub-event -
> 3
2. Steady, or annual or seasonal average loadings required for receiving water
model. Establish qualitative indication of degree of variability with time within the year
or season.
a) Wet-weather loading is relatively predictable based only on annual or
seasonal rainfall totals. This is likely the case when concentrations are
relatively predictable, runoff volumes can be estimated from rainfall, and
event mean concentrations are not strongly correlated with flow. -> 2.1
or
b) Wet-weather loading over the year or season is dominated by a few major
events, as is often the case for combined sewer overflows, and day to day
hydrology must be simulated to obtain a reasonable estimate of annual
totals - 2.2
2.1. Use of annual or seasonal runoff totals is adequate: detailed simulation of wet-
weather quantity is not required.
2.2. Average wet-weathher flows and loading need to be determined for event-
dominated, episodic runoff.
a) Analysis at levels 0-2 -> 2.2.1
or
b) Analysis at Levels 3-4 -> 2.2.2
2.2.1 Level 0-2 analysis for event-dominated, episodic runoff. If, at these Levels, the
receiving water analysis does not involve continuous simulation it is usually not
Ecological
Response
TMDL Modelmg
Components
Chemical
Transport
Sediment
Transport
Runoff
Quality
Waterbody
Hydrodynamics
Kundf
Quantity
Integrated TMDL
Model
4-52
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
appropriate to undertake continuous simulation of the wet-weather hydrology either.
Therefore, use average annual or seasonal volumes, recognizing that this will introduce
uncertainty by not accounting for the role of antecedent moisture conditions. Year to
year variability in precipitation can be represented by frequency distribution or
sensitivity analysis.
2.2.2 Level 3-4 analysis for event-dominated, episodic runoff. More sophisticated
analysis requires a more exact analysis of the frequency distribution of the annual or
seasonal average totals. This may be obtained by forming the time series of averages
from a continuous series of events. Wet-weather quantity should thus be simulated at
the event scale.
3. Continuous, event or sub-event receiving water chemical time step required.
a) Receiving water model requires reasonable maximum load or event mean
loads -> 3.1
or
b) Receiving water model requires detailed pollutograph at the sub-event
scale (hourly to subhourly) - 3.2
3.1 Event based model required; determination of reasonable maximum may require
continuous simulation of events to establish frequency distribution with antecedent
conditions taken into account.
3.2 Continuous model required with hourly to sub-hourly chemical time step as
required by receiving water model.
STEP 7B. WET-WEATHER QUANTITY MODELING - SPATIAL DETAIL/
ROUTING
Ecological
Response
Chemical
Transport
TMDL Modeling
Components
Sediment
Transport
Runoff
Quality
Waterbody
Hydrodynamics
Virtually any wet-weather loading
model is capable of reproducing spatial
variability of sources to some degree. For
instance, the simplest model may be run
several times for each land use area, thus
building up a representation of spatial
variability. Choice of the level of wet-
weather model spatial detail in a
modeling application for TMDL will
reflect factors including 1) level of the analysis; 2) availability of data and financial and
manpower resources; 3) degree of heterogeneity in watershed land use patters; 4) extent
Quantity
Integrated TMDL Model
4-53
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
to which significant wet-weather sources are focused (e.g., entering through CSO outfalls
or drainage swales), or distributed; and 5) spatial resolution required by the receiving
water model. However, because most wet-weather loading models can be used to
express spatial variability, these factors are not generally categorizable determinants for
model selection (although they may be significant in a site-specific context). Therefore,
this section concentrates on whether or not a space-time history of loading is required,
versus simply calculating event (or seasonal) totals. Essentially, this reflects whether any
hydrologic routing of overland, storm sewer, or first order streamflow needs to be
performed to provide input to a receiving Water model. Capabilities of wet-weather
loading models vary widely in this regard.
1- Determine whether receiving water model demands detailed time history or
pollutograph.
a) Analysis is at Level 0 (simple scoping) -> 2
or
b) Analysis is at Level 1 or greater, and receiving water model does not
require intra-event time step -> 3
or
c) Receiving water model chemical time step requires loading model time
step that is intra-event (hourly, sub-hourly) -> 4
2. Level 0 scoping or screening analysis. Only average loading estimates will be
used. No hydraulic routing is required.
3. Level 1 or greater analysis with required loading time step at daily or event scale
or greater.
a) Level 1,2 analysis -> 3.1
b) Level 3,4 analysis -> 3.2
3.1 Level 1 or 2 analysis with loading time step at daily or event scale or greater.
Time dependent routing is not necessary (although it may be used). Runoff coefficient
and curve number methods for hydraulic simulation may be acceptable.
3.2 Level 3 or 4 analysis with loading time step at daily or event scale or greater.
Determine need for routing:
a) Load delivered during event depends on peak or timing of flow within
event -> 4
or
b) Daily loads are required, and events typically last longer than one day, and
significant variability in concentration occurs with time within the event -
> 4
4-54
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
or
c) Neither (a) nor (b) apply -> 3.2.1
3.2.1 Daily/event load calculations without intra-event timing. Water balance (input-
output) and hydrologic routing methods are acceptable (although more sophisticated
routing techniques may also be used.)
4. Intra-event loading time history required by receiving water model, or intra-event
time history required to determine event loading. Routing of wet-weather flows is
required.
a) Hourly time step is sufficient and detailed replication of flood waves or
other momentum changes is not required -> 4.1
or
b) Flood wave propagation is thought to be significant to analysis of flow, but
flow is gradually varying and effects of surcharging and backwaters (in
pipe flow) or flood wave attenuation (in overland and channel flow) are
of minor importance -> 4.2
or
c) Effects of surcharging and backwaters (in pipe flow) or flood wave
attenuation (for overland and channel flow) or other internal momentum
effects must be considered -> 4.3
4.1 Intra-event loading without momentum effects. Can use water balance or
hydrologic routing methods. (More sophisticated routing methods, such as kinematic
and dynamic wave, are also acceptable.)
4.2 Intra-event loading with gradually varying flow. Can use kinematic wave or
dynamic wave routing solutions.
4.3 Intra-event loading with strong internal momentum effects. Desirable to use a
dynamic wave routing solution to provide accurate description.
4-55
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
STEP 8A WET-WEATHER QUALITY MODELING - TEMPORAL RESOLUTION
This step is designed to aid in
determination of the time detail
requirements of the quality component of
a wet-weather loading model - if indeed
the quality aspects are simulated.As wit
the chemical time step in receiving water
models, it is framed in terms of an
applicable maximum time step: the user is
suggested to choose a model which can
implement a temporal resolution of this scale or finer.
As noted above, there will be various instances in which attempting to simulate
wet-weather load quality will not result in much, if any, improvement in predictions over
use of concentrations obtained from monitoring data
1. Determine need for wet-weather quality simulation.
a) Sufficient monitoring database exists so that it is not necessary to simulate
wet-weather quality (i.e., use wet-weather quantity information and site-
specific relation to quality), and it is not desired to predict land use
changes to situations for which monitoring is not available -> 2
or
b) Wet-weather quality simulation is required -> 3
2. No simulation of wet-weather quality is required. Measured quality data are
combined with wet-weather flows for the analysis.
3. Simulation of wet-weather quality is required. Primary determinant of wet-
weather time step is the Level of analysis.
a) Level 0 (screening) analysis -> 3.1
or
b) Level 1-4 analysis -> 3.2
3.1 Level 0 (screening) modeling of wet-weather. Apply annual loading function type
models. An approximate partitioning of the total annual load within the year may be
obtained by assessing the frequency of runoff producing precipitation events above a
certain minimum transport capacity.
3.2 Level 1-4 simulation of wet-weather quality,
a) Level 1,2 analysis -> 3.2.1
or
Ecological
Response
TMDL Modeling
Components
Chemical
Transport
Sediment
Transport
Runoff
Quality
Waterbody
Hydrodynamics
Runoff
Quantity
Integrated TMDL Model
4-56
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
b) Level 3,4 analysis -> 3.2.2
3.2.1 Level 1,2 simulation of wet-weather quality. Loading required by receiving water
analysis is:
a) Steady state, annual or seasonal average loading -> 3.2.1.1
or
b) Event average or reasonable maximum event loading -> 3.2.1.2
3.2.1.1 Level 1,2 simulation of annual or seasonal average wet-weather loading.
Detailed resolution of timing of wet-weather loading not required; simple models with
annual event time scale are adequate; finer resolution models may also be employed.
3.2.1.2 Level 1,2 simulation at event level. Wet-weather model should be able to
simulate typical and extreme events, but not necessarily the detailed time history of
events. Maximum simulation time step is daily or event Event-based or continuous
wet-weather quality models may be used.
3.2.2 Level 3,4 simulation of wet-weather runoff quality, in which a more accurate time
series of loads is required. Continuous or probabilistic simulation should be employed.
The time step limit established for simulating loading to receiving waterbody is:
a) Daily or event -> 3.2.2.1
or
b) Sub-event, hourly or sub-hourly -> 3.2.2.2
3.2.2.1 Continuous time series of wet-weather event loads is needed, but not intra-
event timing of loads. Use maximum wet-weather quality simulation time step at the
daily or event level.
3.2.2.2 Continuous time series of wet-weather loads is needed with intra-event
loading history. User should evaluate whether to use a daily-event time scale model
(with flow-weighted partitioning to the desired loading time scale) or a sub-event time
scale model.
4-57
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
STEP 8B. WET-WEATHER QUALITY MODELING - SPATIAL DETAIL
As with quantity modeling, most
wet-weather load models can represent
spatial heterogeneity of sources by
repeated application to subareas, if more
sophisticated methods are not built in.
The question addressed here is whether
the model should address pollutant
routing: i.e., the space-time distribution of
pollutant loads, rather than just the
loading history predicted by transport by advective flows. This may be important when
the rate of transport of the pollutant differs significantly from the rate of flow of water,
or where reactions are taking place within the wet-weather flow.
1. Evaluate based on wet-weather quality simulation temporal resolution.
a) Wet-weather quality temporal resolution is the event scale or greater, and
it is not necessary to simulate the effects of in-transit treatment -> 2
or
b) Wet-weather quality temporal resolution is less than event length, or
detailed analysis required of in-transit treatment effects ->3
2. Detailed quality routing capabilities are not needed from the wet-weather load
model.
i
3. Intra-event wet-weather quality simulation desired. A sensitivity analysis may
be performed to determine if it makes any significant difference to predictions whether
the exact arrival timing of component parts of the event load, beyond that predicted by
simple advection, is considered. This should usually only be necessary at Levels 3-4 of
analysis.
a) Sensitivity analysis indicates non-advective intra-event timing of
components of load should be considered: Use a model with wet-weather
quality routing capabilities.
or
b) Sensitivity analysis indicates non-advective intra-event timing of
components can be ignored: Wet-weather quality routing capabilities are
not required in the model (as distinct from hydraulic routing capabilities).
Ecological
Response
TMDL Modeling
Components
Chemical
Transport
Sediment
Transport
Waterbody
Hydrodynamics
Runoff
Quantity
Integrated TMDL Model
4-58
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
STEP 9. ASSEMBLE AND REVIEW MODEL IDENTIFICATION CRITERIA
Ecological
Response
TMDL
Com
Modeling
>onents
Chemical
Transport
Sediment
Transport
Runoff
Quality
Waterbody
Hydrodynamics
Runoff
Quantity
rated TMDjP
L. if J-1a1
VfOuCi
Step 0 provides the TMDL
developer with a table of important
physical, regulatory and functional
characteristics of the system to be
modeled. Steps 1 through 8 then yield
suggestions regarding the characteristics
of models which may be appropriate for
TMDL analysis of the system. These can
be assembled to provide the basis for
model identification. (Note that Step 9 is not a decision tree; the following items should
be addressed in sequence)
9.1. Form Table of Model Identification Criteria. The user should compile the
results of all these steps in a tabular form (see, for example, Table 4-3). This table will
summarize the types of models which might be employed for TMDL analysis, and some
of their important functional requirements.
Table 4-3. Technical Criteria Summary Form for Model Selection
Model
Component
Is a
model
needed?
Physical
Domain
Pollutant
or Impacts
Addressed
Time Scale
of
Simulation
Flow
Routing
Capabilities
Quality Routing
&
Transformations
Ecological
Response
...
¦
: \ ' '¦
•• gsj
1 Hill 1 If
Receiving
Water Quantity
Receiving
Water Quality
•
Sediment
• •
NPS Quantity
,
NPS Quality
9.2. Review Model Identification Criteria Table. Once the table is assembled,
4-59
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
it should be subjected to a thorough review. It is important to remember that Steps 1
through 7 provide only generic recommendations. These have been formulated to cover
a wide range of circumstances; however, they may not be appropriate for the site-specific
conditions of the problem under study. Two important questions that should be
addressed in the review are:
• Do the model identification criteria make sense in terms of the known
conditions of the waterbody and pollutant under study? To the extent that
they do not, the derisic process should be revisited and alternative
specifications examined and justified.
• Do the model identification criteria make sense in terms of the proposed
level of effort and degree of accuracy required for the evaluation? If the
analysis is thought to be too simple and uncertain, this may indicate that
the decision trees should be revisited with the choice of the next higher
Level of analysis. Where the analysis is thought to be too complex or
expensive, it may sometimes be possible to move to a lower, more
approximate Level of analysis, with a correspondingly larger MOS.
However, in this case the user should also examine whether it is feasible
to perform the TMDL at all within available resource limitations.
9.3. Analysis of Candidate Models. After the model identification criteria have
been revised, a list of candidate models can be formed. The different model components
(e.g., wet-weather loading quality and receiving water quality) may be addressed by
separate models, or, if available, within single modeling packages. The candidate
models should possess technical characteristics which match those identified in the
model identification criteria table. Tabular guides to many Federal agency supported
simulation models were provided in Chapter 3. Additional models available to the user
are readily added to the list
9.4. Model Selection. The previous step should identify available simulation
models, or analysis methods, which meet the technical and regulatory needs of the site
TMDL analysis (see section 4.2), with some recognition of user functional and
operational needs, via choice of an appropriate Level of analysis. Where multiple
candidate models are identified, the final choice of simulation models can further reflect
user functional and operational needs, including cost and time requirements for the
implementation, ease of use, and familiarity of staff with the models.
4-60
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
References. Chapter 4
Ambrose, R.B., J.L. Martin and J.F. Paul. 1990. Technical Guidance Manual for
Performing Waste Load Allocations, Book HI: Estuaries, Part 1: Estuaries and
Waste Load Allocation Models. EPA/823/R-92-002. U.S. EPA OW/OST.
Bowie, G.L. et al. 1985. Rates, Constants, and Kinetics Formulations in Surface Water
Quality Modeling (2nd Edition). EPA/600/3-85/040, Environmental Research
Laboratory, Athens, GA
Donigian, A.S. Jr. and W. C. Huber. 1991. Modeling of Nonpoint Source Water Quality
in Urban and Non-urban Areas. EPA/600/3-91/039. ERL Athens, GA
Dunne, T. and L.B. Leopold. 1978. Water in Environmental Planning. W. H. Freeman
and Co., San Francisco.
Eschenroeder, A. 1983. The role of multimedia fate models in chemical risk analysis. In:
Fate of Chemicals in the Environment. ACS Symposium Series 225. American
Chemical Society, Washington, DC
Fausch, K.D., C.L. Hawks and M.G. Parsons. 1988. Models That Predict Standing Crop
of Stream Fish from Habitat Variables (1950-1985). General Technical Report
PNW-GTR-213. USDA, Forest Service, Pacific Northwest Research Station,
Portland, OR
Fischer, H.B., E.J. List, R.C.Y. Koh, J. Imberger and N.H. Brooks. 1979. Mixing in Inland
and Coastal Waters. Academic Press, Orlando.
Freedman, P.L., D.W. Dilks and B.A. Monson. 1992. Technical Guidance Manual for
Performing Waste Load Allocations, Book HI: Estuaries, Part 4: Critical Review of
Coastal Embayment and Estuarine Waste Load Allocation Modeling. EPA-823-R-
92-005. Office of Water.
Hydroqual, Inc. 1986. Technical Guidance Manual for Performing Wasteload Allocations,
Book IV, Lakes, Reservoirs and Impoundments, Chapter 3, Toxic Substances
Impact. EPA-440/4-87-002.
Leopold, L.B., M.G. Wolman and J.P. Miller. 1964. Fluvial Processes in Geomorphology.
W.H. Freeman, San Francisco.
LimnoTech, Inc. 1985. Dynamic Toxics Waste Load Allocation Model (DYNTOX): User's
Manual. Office of Water Regulations and Standards.
4-61
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Mandni, J.L., P. A. Mangarella, G. Kaufman and E. D. Driscoll. 1983. Technical Guidance
Manual for Performing Waste Load Allocations, Book IV, Lakes and
Impoundments, Chapter 2, Nutrient/Eutrophication Impacts. EPA-440/4-84-029.
OWRS Monitoring and Data Support Division, U.S. EPA.
Mao, K. 1992. How to select a computer model for storm water management. Pollution
Engineering, Oct. 1,1992, pp. 60-64.
McKeon, T.J. and J.J. Segna. 1987. Selection Criteria for Mathematical Models Used in
Exposure Assessments: Surface Water Models. EPA/600/8-87/042. Exposure
Assessment Group, Office of Health and Environmental Assessment, U.S. EPA.
Rosgen, D.L. 1993. Classification of natural rivers, (submitted to Catena)
Schnoor, J.L. 1985. Modeling chemical transport in lakes, rivers, and estuarine systems.
In: Neely, W.B. and G.E. Blau, eds., Environmental Exposure from Chemicals, vol.
2. CRC Press, Boca Raton, FL
U.S. EPA. 1985. Guidelines for Deriving Numerical National Water Quality Criteria for
the Protection of Aquatic Organisms. NTIS PB85-227049.
U.S. EPA. 1986. Technical Guidance Manual for Performing Wasteload Allocation, Book
VI, Design Conditions, Chapter I, Stream Design Flow for Steady-State Modeling.
EPA/440/4-87-004.
U.S. EPA. 1989. Resolution on Use of Mathematical Models by EPA for Regulatory
Assessment and Decisionmaking. EPA-SAB-EEC-89-012. Environmental
Engineering Committee, Science Advisory Board.
U.S. EPA. 1990. Biological Criteria, National Program Guidance for Surface Waters.
EPA-440/5-90-004. Office of Water.
U.S. EPA. 1991. Technical Support Document for Water Quality-based Toxics Control.
EPA/440/4-91-001. OWEP/OWRS.
4-62
-------
The Cadmus Group, Inc.
DRAFT
June 3, 1993
Chapter V. Monitoring Plans and Impact Assessment for TMDLs
Purpose: This chapter addresses potential needs for monitoring to support
modeling and to track potential benefits from implementing new wastewater control
and/or watershed management practices. It introduces necessary considerations for
designing a monitoring program to provided results adequate for addressing the
program's general monitoring goal and specific monitoring objective. Chapter 6
introduces sampling techniques, primarily focusing on the general mechanics and
potential problems with sampling. It also provides guidance on some techniques
available for summarizing and assessing collected data. Particular attention in both
chapters is paid to special considerations for sampling related to wet weather TMDL
modeling efforts.
The chapter focuses on considerations important for developing monitoring plans.
It introduces existing documents that contain information useful in designing monitoring
plans, especially those applicable to wet-weather TMDL studies. Distinctions between
monitoring goals to support modeling efforts, monitoring goals to track benefits of
control, and management efforts are discussed, as are considerations for defining specific
monitoring objectives. Also, special considerations are introduced regarding impacts and
associated monitoring requirements for various types of waterbodies and land-uses.
5.1 Interaction of Monitoring and Modeling for TMDLs
An appropriate modeling strategy should be designed to provide answers
appropriate to the TMDL process, within the constraints of the physical characteristics
of the waterbody, pollutants and impacts under study, available resources, and other
functional and operational needs of die TMDL developer. In most cases it will not be
possible to specify all the characteristics of the system at the beginning of the process,
when data may be limited. (This is, indeed, why a model selection strategy is required).
Instead, TMDL developers will need to refine the goals and requirements of modeling
as they gain additional knowledge through system characterization and monitoring.
Similarly, TMDL developers will likely need to adjust monitoring plans as the
characterization and assessment effort proceeds. The modeling strategy will help
determine the calibration data requirements from the monitoring program, and model
results may suggest critical areas where monitoring will be most effective.
These observations indicate that the development of a final modeling strategy
(and, indeed, determination of the need to employ sophisticated modeling) is an iterative
process, which will involve close interplay between modeling and monitoring. For
instance, a TMDL developer might use a simple screening model for initial scoping of
5-1
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
the problem. This can aid in designing the monitoring plan, which in turn can lead to
1) determination of the need for more sophisticated modeling, and 2) the formulation of
a targeted modeling strategy for the completion of the TMDL assessment.
5.2 Existing Guidance for Information on Monitoring Designs and Sampling
Procedures
A number of existing guidance documents include information valuable to the
design of monitoring programs and on selection and use of appropriate sampling
procedures. Table 5-1 presents an list of nearly 25 documents produced or sponsored
by EPA. Table 5-2 presents a selection of other guidance documents available from non-
EPA sources that also contain much valuable information useful to the design and
implementation of monitoring programs. While these lists are extensive, they are not
comprehensive lists of available guidance. Many of the these documents were used to
help develop some of the discussion included in this chapter, and they should be
consulted whefe users of this document require additional information on the topics
discussed.
5-2
-------
The Cadmus Group, Inc.
DRAFT
June 3, 1993
Table 5-1. Existing EPA guidance on monitoring designs and sampling procedures
Guidance for Water Quality-based Decisions: The TMDL Process (U.S. EPA, 1991a) - Introduces the steps of
the TMDL process, including general requirements for modeling and monitoring.
Water Quality Standards Handbook (U.S. EPA, 19b ja) - Provides general guidance on water body surveys
and assessments for conducting use attainability analyses and guidelines for developing site-specific water
quality criteria.
Technical Support Manual: Waterbody Surveys and Assessments for Conducting Use Attainability Analyses
[streams and rivers] (U5. EPA, 1983b) - Detailed guidance on conducting physical, chemical, and
biological evaluations of stream and rivers, and appropriate methods for summarizing and assessing
collected data.
Technical Support Manual: Waterbody Surveys and Assessments for Conducting Use Attainability Analyses.
Volume II: Estuarine Systems (U.S. EPA, 1984a) - Introduction to the physical dynamics of estuarine
systems and the characteristics of plant and animal communities; detailed guidance on methods to
characterize physical, chemical, and biological conditions, ami on approaches to synthesize and interpret
collected data.
Technical Support Manual: Waterbody Surveys and Assessments for Conducting Use Attainability Analyses.
Volume III: lake Systems (US. EPA, 1984b) - Review of physical, chemical, ami biological characteristics
of lakes and reservoirs; detailed introduction to sampling methods and methods to summarize and assess
collected data.
Technical Guidance Manual for Performing Waste Load Allocations. Book II. Stream and Riven. Chapter 1,
Biochemical Oxygen Demand/Dissolve Oxygen. (Driscoll et al., 1983) - Introduction to the assessment of BOD
and IX) in flowing water systems; minimum guidance on sampling and monitoring.
Technical Guidance Manual for Performing Waste Load Allocations. Book II. Stream and Rivers. Chapter 2,
Streams and Rivers: Nutrient/Eutrvphication Impacts (U.S. EPA, 1983c) - Introduction to evaluations of
nutrient and eutrophication problems in flowing water systems; emphasis on assessment of phytoplankton
appropriate only for large, slowly flowing rivers; guidance on minimal sampling requirements.
Technical Guidance Manual for Performing Waste Load Allocations. Book IE. Estuaries. Part 1: Application of
Estuarine Waste load Allocation Models (Martin et al., 1990) - Provides monitoring protocols for calibration
and validation of estuarine models, including types of data, frequency of collection, spatial coverage, and
quality assurance.
Technical Guidance Manual for Performing Waste Load Allocations. Book IV. Lakes and Impoundments. Chapter
3: Toxic Substances Impact (Hydroqual, Inc., 1986) - Introduction to problem of toxic substances in lakes
aiwi reservoirs; presents data requirements for models, including minimum sampling frequencies,
problems of time scales, needs for analysis of dissolved or total, concentrations, sample handling,
preservation, and documentation; does not specifically address WLA or TMDL processes.
5-3
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Table 5-1. (continued)
Handbook Stream Sampling for Waste Load Allocations Applications (Mills et al., 1986) - Provides guidance
for designing stream surveys to support modeling applications for WLAs; includes potential sampling
requirements for meteorological data, water quality data, times, frequencies, etc.
Guidance for State Water Monitoring and Wasteload Allocation Programs (U.S. EPA, 1985) - Provides general
introduction on determining needs for monitoring program designs, monitoring for water quality based
controls, monitoring for compliance and enforcement, water quality assessments, quality assurance, data
reporting, EPA's monitoring strategy, and monitoring checklists.
Technical Support Document far Water Quality-based Toms Control (US. EPA, 1991b) - Introduces needs for
biological criteria, biological assessments, and biological surveys; monitoring needs in establishing permit
limits.
NPDES Stormwater Sampling Guidance Document (US. EPA, 1992a) - Focuses on technical aspects of
stormwater sampling, particularly to meet requirements for NPDES permits; includes extensive discussion
of fundamentals of sampling for flow measurements and water quality evaluations.
Watershed Monitoring and Reporting for Section 319 National Monitoring Program Projects (U.S. EPA, 1991c) -
Introduces water quality monitoring and assessment approaches under the 319 Program/ including the
design of monitoring programs; sampling station location; physical, chemical, and biological monitoring
parameters; data analysis; and reporting requirements.
Guidance Specifying Management Measures for Sources ofNonpomt Pollution in Coastal Waters (US. EPA, 1993)
-Includes an extensive chapter on monitoring ami tracking techniques to accompany management
measures.
Developing Nonpomt Source Load Allocations for TMDLs - A Quick Reference Guide (Tetra Tech, Inc., 1992) -
Includes a brief discussions on the role of water quality monitoring and follow-up monitoring; highlights
additional sources for NPS monitoring guidance.
Biological Field and Laboratory Methods for Measuring the Quality of Surface Waters and Effluents Weber, 1973)
-The classic EPA guide to sampling methods and data analytical approaches for plankton, periphyton,
macrophytes, macroinvertebrates, fish, and toxicity bioassays.
Macroinvertebrate Field and Laboratory Methods for Evaluating the Biological Integrity of Surface Waters OQemm
et al., 1990) - Provides thorough guidance on monitoring designs, sampling procedures, and analysis
approaches for macroinvertebrate communities inhabiting aquatic environments.
Rapid bioassessment Protocols for Use in Streams and Rivers - Benthic Macroinvertebrates and Fish (Flafkin et
al., 1989) - Presents five specific protocols, including detailed guidance on sampling and data analysis
methods, for assessing biological conditions in streams and rivers.
Microbial Methods for Monitoring the Environment - Water and Wastes (Bordner and Winter, 1978) - Presents
detailed guidance on sample collection and analytical methods for microbial samples.
5-4
-------
The Cadmus Group, Inc.
DRAFT
June 3, 1993
Table 5-1. (continued)
Monitoring Lake and Reservoir Restoration (Wedepohl et al., 1990) - Presents example monitoring designs
and sampling methods for lake, lake tributary, and watershed monitoring programs under Clean Lakes
Program projects.
Monitoring Guidelines to Evaluate Effects of Forestry Activities on Streams in the Pacific Northwest and Alaska
(MacDonald et al., 1991) - Provides detailed guidance on designing monitoring programs and on assessing
monitored physical, chemical, and biological variables; includes a discussion on using the TMDL process
within these assessments.
Ecological Assessments of Hazardous Waste-Sites: A Field and Laboratory Reference Document (Warren-Hicks
et al., 1989) - Introduces uses and development of ecological effects endpoints, assessment strategies and
approaches, field sampling designs, toxicity testing, biomarkers, and field assessment procedures for
aquatic and other environments.
Table 5-2. Additional guidance on monitoring designs and sampling procedures
Design of Water Quality Monitoring Systems (Ward et al., 1990) - Excellent source of guidance to monitoring
plan design, includes the role of statistics, quantifying information expectations, data analysis, and
network design.
Sampling Design and Statistical Methods for Environmental Biologists (Green, 1979) - Includes 'Ten Principals"
of design, "optimal" impact study designs, selection of monitoring variables, and presentation of results.
Field Methods and Statistical Analyses for Monitoring Small Salmonid Stream (Annor et al., 1983) - USDI Fish
and Wildlife Service guidance on assessing land use impacts, selecting and applying physical, chemical,
and biological assessment methods, and statistical analyses of aquatic environmental monitoring data.
Methods for the Assessment and Prediction of Mineral Mining Impacts on Aquatic Communities: A Review and
Analysis (Mason, 1978) - Proceedings from a USDI Fish and Wildlife Service workshop that produced
detailed guidance on deigning impact studies, analyzing collected data, and sampling and assessing
bacteria, algae, zooplankton, benthic maaoinvertebrates, fish, and other aquatic-related taxa.
Methods for Evaluating Riparian Habitats xvith Applications to Management (Platts et al., 1987) - USDA Forest
Service document that provides extensive guidance on collecting Hat* and assessing riparian areas,
emphasizing Western streams.
Methods for Evaluating Stream, Riparian, and Biotic Conditions (Platte et al., 1983) - USDA Forest Service
guidance on selecting study sites, developing sampling designs, using sampling transects, and assessing
riparian, fish population, and benthic macroinvertebrate conditions in stream studies.
5-5
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Table 5-2. (continued)
Estimating Total Fish Abundance and Total Habitat Area in Small Streams Based on Visual Estimation Methods
(Hankin and Reeves, 1988) - Benchmark paper that provides habitat characterization methods that provide
a basis for many of the more easily used approaches for monitoring stream habitats.
Fisheries Habitat Surveys Handbook (USDA Forest Service, 1989a) - One of several handbooks developed by
the USDA Forest Service's Regions including intensive procedures to standardize survey and data
collection procedures used to assess stream and lake habitats on Forest Service lands.
A Basin-Wide Stream Inventory Process Using Habitat Type Classifications for Resident Sahnomds (USDA Forest
Service, 1989b) - An example of broadly used alternative procedures used by individual National Forests,
which are based on the Hankin and Reeves (1988) method, to characterize and assess stream habitat
5.3 Defining General Monitoring Program Goals
Most monitoring programs collect data to generate information. The distinction
between data and information is important Put simply/ data are measurements or
observations made on a system (e.g., a wastewater stream or an ecological system)/ while
information is knowledge with which one can make decisions. Poorly thought-out or
incomplete monitoring programs often produce "data rich" but "information poor" results
(Ward et al., 1990). For example, samples may have been taken at locations selected
solely because of convenient access, at times when a staff member was near the area or
when the weather was "fair." Such programs often proceed with only vague notions of
what data might be needed and without a good idea of how to analyze the data that are
obtained. When this approach continues over extended periods the results are often data
rich and information poor, because monitoring did not proceed with clear goals,
objectives, or designs.
Table 5-3 shows a series of useful steps to consider when developing monitoring
programs. In the first step, TMDL developers should identify die goal(s) of monitoring.
In the context of this guidance, monitoring generally would be conducted to address one
or both of two general goals. The first of these goals is to support identification model
calibration, and validation. Here the monitoring program can be strictly a data-collection
effort, and the program ends when its collected data are transferred to the modelers.
This is an example of monitoring that has a "data end-point goal" (Table 5-3). Often,
however, monitoring aims to achieve an "information goal" (Table 5-3). Thus, the second
common monitoring goal is to track and assess (1) potential benefits from implementing
new treatment, control, and management strategies, or (2) potential impacts from new
discharges. This goal requires analysts to extrapolate or transform monitoring data into
5-6
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
information of known reliability about the characteristics of the system. In other words,
although one intermediate end point of monitoring design and sampling collection is a
set of data, most users are interested in the information conveyed by the data and not
the data per se (except as noted above). They want answers to such questions as:
• How do the ambient conditions of this water body compare to those of
other, similar water bodies in the region?
• How has the new control program improved conditions for aquatic life in
the receiving water?
• How much precipitation-produced runoff will result in NPS runoff and
discharges to the receiving waters?
• How does the discharge volume change when the runoff is due to
snowmelt?
• What are the average ambient conditions in the river?
• How does a waste water discharge of volume "X" alter ambient water
quality conditions in the receiving water?
To address such questions and to address the kinds of information goals shown
in the above examples requires that monitoring designs also include plans for data
management and analyses to derive available information from the collected data.
5-7
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Table 5-3. General Steps for Monitoring Program Development
1. Identify Monitoring Goal(s) (Purpose)
1.1. Data end-point goal
1.1.1. Gather data for system model identification and selection
1.1.2. Gather data for system model calibration and verification
12. Information goal
12.1. Establish characteristics of system variables (e.g., for water quality compliance
monitoring)
12.2. Assess ecological impacts to biological populations and communities inhabiting
receiving waters
2. Define Monitoring Objective(s)
2.1. Characterize effluent
22 Define system components) or parameters) (i.e., system variables or system locations)
for which monitoring data are needed, the kind(s) of data needed, and minimum accuracy
and precision of data required to meet the information needs under each monitoring
objective(s)
2.3. Define the limits of variability in system conditions (e.g., time of day, season, section of
runoff hydrograph) to be characterized under each monitoring objective
2.4 Define Data Quality Objectives
2.5 Define how data will be used, what comparisons made, what procedures used
3. Design Sampling Program
3.1. Data end-point objectives
3.1.1. Schedule sample and data collection periods, frequencies, and locations to
correspond to requirements, established under eaph monitoring objective
3.1.2. Identify and select field sampling, data collection techniques, and laboratory
analytical techniques that will provide data of sufficient precision and accuracy
to achieve each monitoring objective
5-8
-------
The Cadmus Group, Inc. DRAFT June 3,1993
Table 5-3. (continued)
32. Information objective
3.2.1. Identify and select statistical method(s) capable of providing information required
to meet each monitoring objective
32.2. Define minimum data requirements (data quality and quantity) that allow each
selected statistical method(s) to be used to produce the information required to
meet each monitoring objective
32.3. Schedule sample and data collection periods, frequencies/ and locations to
correspond to requirements established under each monitoring objective
32.4. Identify and select field sampling and data collection techniques and laboratory
analytical techniques that will provide data of sufficient precision and accuracy
to achieve each monitoring objective
Monitoring Program Implementation (see Table 5-4)
4.1. Sample and field data collection
42. Laboratory analysis
43. Data management
4.4. Data analysis
45. Reporting
4.6. Information use
5-9
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Table 5-4. Checklist of Considerations for Documenting Monitoring Program
Designs and Implementation (expanded from Ward et aL, 1990)
Sample and Field Data Collection
Pre-Sampling Preparations
Personnel selection and responsibility identification
_ Personnel safety training and health care verification (first aid training,
CFR, wet-weather training, safety guides, vehicle safety, vaccinations
current, etc.)
Site access prepared and legal consents obtained
Scientific sampling or collecting permits needed
Formats for field sampling logs and diaries
Equipment availability, acquisition, and maintenance
Sample collection schedule (random? regular? same-time-of-day?)
Preparation of pre-sampling checklist
Sampling procedures
_ Procedures documentation
Staff qualifications and training
__ Pre-sampling preparation (e.g., purging sample lines, instrument
calibration)
Sampling protocols
_ Quality control procedures (equipment checks, replicates, splits, etc.)
__ Required sample containers
_ Sample numbers and labeling
__ Sample preservation (e.g, "on ice" or chemical preservative)
Sample transport (delivery to laboratory needed?)
Sample storage requirements met
Sample tracking and chain-of-custody procedures
Field measurements
Field log and diary entries
Sample custody and audit records
Post-Sample Follow Up
_ Filing sample logs and diaries
_ Equipment cleaning and maintenance
__ Proper disposal of chemical wastes
__ Review documentation and audit reports
5-10
-------
The Cadmus Group, Inc.
DRAFT
June 3/ 1993
Table 5-4. (continued)
Laboratory Analysis
Pre-Sample Analysis Preparations
Verify use of proper analytical methods
Analyses scheduling
Verify sample number
Define a recording system for sample results
Apply a system to track each sample through the lab
Equipment maintenance and calibration
Quality control solutions on hand
Sample Analysis
Sample analysis methods and protocols
Use of reference samples, duplicates, blanks, etc.
Quality control and quality assurance compliance
Sample archiving
Proper disposal of chemical wastes
Full documentation in bench sheets
Data Record Verification
Coding sheets, data loggers
Data verification procedures and compliance with project plan
Analysis of splits are within data quality objectives
Assignment of data quality indicators and explanations
Data Management
Selection of appropriate hardware and software
_ Data entry practices and data validation (e.g., entry range limits, duplicate
entry cheddng)
Data tracking
Characteristics of data achieving system
Data exchange protocols
General data availability
5-11
-------
The Cadmus Group, Inc.
DRAFT
June 3, 1993
Table 5-4. (continued)
Data Analysis
Software selection
Handling missing data and non-detects
Identification and use of data outliers
Graphical procedures (e.g., scatter plots, notched-box and whisker)
Parametric statistical procedures
Non-parametric statistical procedures
Trend analysis procedures
Multivariate procedures
Quality control checks on statistical analyses
Reporting
Scheduled reports—timing, frequency, and lag times following sampling
Report contents and formats
Planned tables and graphics
Report sign-off responsibility(ies)
Report distribution recipients and availability
Use of paper and electronic formats
Presentations
Information Use
_ Identification and application of decision or trigger values, resulting action
_ Implementation of construction, control, and/or monitoring design
alternatives
Public release procedures
General
Contingencies
Follow-up procedures
5-12
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
5.4 Defining Specific Monitoring Program Objectives
With the goal(s) for monitoring defined, the next task when designing monitoring
programs is to define the specific monitoring objective(s) (Table 5-3). Monitoring
program objectives refine the broad goal(s) established for the monitoring program by,
most simply, formally defining the information needs for why monitoring is to be
undertaken. These objectives delimit the specific needs for monitoring data collection
and specifies how these data will be used after collection, for example, by management
or by regulatory agencies. They also provide limits for information expectations. With
many monitoring programs, sampling and data collection efforts may require a phased
implementation to address progressively more specific and detailed unknowns
(objectives).
An important step during the process of defining specific monitoring objects is
to compile all available relevant data for the system of concern. Then, at least a
preliminary assessment of these data should be completed to determine kind and detail
of information already available for the system. This often serves to focus needs for
additional monitoring by step clearly establishing what is already known and what more
needs to be learned about the system. For example, available information for the system
should be compiled that, to the extent possible, identify, locate, and characterize:
• Point source discharges,
• Known and suspected nonpoint source discharges,
• Sites that have extensive existing data bases that warrant continued or
intensified monitoring to provide long-term data sets useful to the goals
of the planned monitoring program,
• Ambient receiving flow regimes and associated water qualities,
• Potentially affected sensitive receiving water, riparian, and wetland
habitats, and
• Key indicator species, species of special concern, and species listed as
Federal or State threatened or endangered species or are candidates for
such listing.
Defining specific monitoring objectives requires focusing on realistic targets for
specific monitoring activities. In this step, the program designer must identify the
system variables and locations for which new monitoring data are needed. This step
also requires defining the quality (accuracy and precision) of data needed to meet the
5-13
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
monitoring objective. Monitoring programs provide sample data from limited locations
and times across a range of conditions. Hence, selected sampling locations and times
must provide "representative" samples under the conditions of concern to meet the
monitoring objective. Each monitoring objective should specify the limits of system
variation to be covered by the monitoring samples. Section 5.5 introduces implications
of effluent, environmental, and sampling variability in defining monitoring designs.
Definitions of monitoring objectives should be as focused as possible. They
generally should be clear and simple enough to allow fulfillment of the objectives
without the needs for inventive or otherwise unproven sampling, analysis, or assessment
methods. To achieve complex monitoring goals, program designers can group several
monitoring objectives and address them through coordinated efforts or through a series
of linked studies to address each objective until achieving the overall monitoring goal.
Various factors implicit in the these issues fall into a generally larger category of
considerations required for developing data quality objectives (DQOs). In general, EPA's
policy is that DQOs should be developed for all data collection efforts that require or
result in substantial commitment of resources. It is not uncommon to include DQOs as
part of the quality assurance plans for a project
DQOs are qualitative and quantitative statements developed by a team of
technical experts (including someone with statistical expertise), program staff, managers,
decision makers, and other data users to specify the quality of data needed to support
specific technical decisions or regulatory actions. Often, during this development
process, decision makers will describe information needs, reasons for the need, how the
information will be vised, and specifications of any time or resource constraints on data
collection. Next, technical staff and decision makers interact to detail specifications
required for the problem assessment and any technical constraints possibly imposed on
data-collection activities. Then, alternative approaches for data collection are defined
and the approaches) to be used selected. A dear definition of data objectives and
selection of correct data-collection methods help ensure successful study completion.
U.S. EPA (1985) describes the process in detail for developing DQOs; and Plafkin et al.
(1989) also summarizes this material and its importance in quality assurance and quality
control (QA/QC) procedures. We also include additional discussion on QA/QC
considerations in Section 6.5.
5.5 Specifying Sampling Designs: Sample Sizes, Frequencies, and Locations
After clearly defining the monitoring objectives, the next step is to design the
monitoring program (Table 5-3), sometimes also described as the monitoring system
(e.g., Ward et al., 1990). Thus, monitoring program designers must not only define
5-14
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
clearly the monitoring's short-term and (often) long-term objectives, but also the details
of the monitoring process. A well-defined monitoring design should address:
• What are the objectives of monitoring?
• When will monitoring begin?
• What environmental (samp e) variables will be monitored?
• What methods will be used to analyze monitored variables?
• Where and how often will monitoring occur?
• How will collected data be managed and analyzed?
• What information will satisfy the program's goals, thus allow this phase
of monitoring and, in some cases, the entire monitoring program
eventually to end?
Monitoring objectives may require sampling designs of varying degrees of
complexity, for instance:
• Point-in-time, single event, samples to broadly describe the general
character of the sampled variable.
• An intensive, short-term series of collected samples over a predetermined
period to detail patterns of change in monitoring variables associated with
particular events, e.g., runoff from nonpoint pollution sources or a
combined sewer overflow. Sample collections for such studies may occur
at, for example, 5-minute, hour, or day intervals, depending on the
information needs of the monitoring objective.
• Long-term samples collected at regular intervals (e.g., weekly/ monthly,
quarterly, or annually ) to establish ambient or background conditions, or
to assess, for example, general long-term trends of change or seasonal time
series patterns in the monitored variables.
• Reference site studies to provide data for monitored variables against
which results from other monitoring sites can be compared to judge
relative changes in the variables between sampling dates (see Section 7.3).
• Near-field studies to sample and assess receiving waters within the
5-15
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
immediate mixing zone of discharges or runoff. These studies can be used,
for example, to address monitoring objectives regarding possible impacts
due to short-term, acute toxicity or long-term alterations to the habitat.
• Far-field studies to sample and assess receiving waters not within the
mixing zone, but potentially affected by impacts that are delayed due to
longer travel times required for discharged materials to reach these areas.
Studies at such site may include assessments of possible chronic toxicity
or nutrient enrichment effects.
The design of monitoring plans can involve a broad array of technical disciplines.
Previous experience with these diverse problems can greatly aid in developing
appropriate study designs. For example, evaluating and assessing loadings from point
and nonpoint source discharges and their potential impacts—both on beneficial uses and
on receiving water ecosystems—can involve consideration of.
• precipitation patterns,
• watershed hydrology and runoff coefficients,
• collection system hydraulics,
• surface water hydrology,
• water chemistry,
• aquatic toxicology,
• biological assessments,
• system modeling, and
• related areas of science and engineering.
When the monitoring program's goal is only, for example, to provide data for use
in modeling applications/ the design must match sample and data collection periods,
frequencies, and locations to the scheduling requirements established under the
monitoring objectives. Similarly, field sample and data collection techniques and
laboratory analytical methods must be able to provide data that meet die quality
requirements specified by the monitoring objectives before the monitoring program is
ready to implement
5-16
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
But, when the monitoring goal includes information objectives, monitoring
program designers must identify and select statistical, graphical, or other data analysis
techniques that can explicitly address the questions posed by each of these objectives.
Data analysis techniques often differ in their assumptions about the data (e.g., the
distribution of population or sample data) and in their sensitivity to variations in the
data. Also, some techniques require larger sample sizes or more frequent sample
collections than others. Therefore, to collect as much data as efficiently as possible, the
sampling program's design should clearly identify the data analysis methods to be used.
This will help ensure that data collection matches the data needs of the analysis
methods. Definition of optimal monitoring program design must gather sufficient data
to ensure that the monitoring objectives can be met, without collecting data in excess of
the information needs. When first designing a monitoring program or when addressing
new or unusual data problems, TMDL developers may wish to consult statisticians or
other experts in the design of field monitoring programs.
As progressively more data are collected during implementation of monitoring
programs, the relative amount of new information provided may progressively
decreases. One problem in designing an optimal monitoring program is to determine
an information/cost effective break-point for data collection efforts. This generally
involves determining the minimum number of locations and times to be sampled, and
minimum samples to be collected during monitoring to provide the information needed
about the monitored system at the minimum level of acceptability (e.g., statistical
significance).
The number and placement of sampling locations depends on the length of
flowing water system or the extent of the lake, estuary, wetland, or other standing water
system for which the monitoring characterization is needed. It also depends on the
horizontal and vertical variability occurring throughout the area needed characterization
and the degree or resolution of characterization required.
In total, establishing sampling locations must be based on objectives of the study
and must be done considering various scales (Ward et al., 1990). First, at the macro level
the sampling stations locations define the river reach, or lake or estuary area
characterized through the sampling. These locations should directly relate to the needs
defined by the monitoring plan objectives. The micro scale relates to the specific
locations sampled at each sampling station. If the study objective is to characterized the
aquatic community of a stream reach, then the samples should be collected from all
habitat types found in that stream reach (riffles, pools, runs, side channels, etc.). If,
instead, the objective is to define baseline stream conditions against which future
possible changes in the system could be assessed, then sampling locations may be, for
example, limited to riffle areas occurring in 0.75 to 1.5 feet of water. Most importantly,
care must be taken in developing monitoring and sampling plan to avoid sampling the
5-17
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
"wrong water." That is, collecting samples that do not the supply needed information
and cannot be used to address the sampling objectives.
Determining the intervals and frequencies for sampling at monitoring locations
also depends on monitoring objectives. For statistical reasons, it would generally be best
to collect samples at random intervals, but practical constraints often require regularly
spaced intervals. Defining the frequency of sampling (e.g., time between samplings) can
also depend on die rate of natural or other change inherent in the variable sampled.
Water chemistry can change in minutes and some water quality models need calibration
and verification data collected at 5-minute intervals, while other monitoring objectives
can be satisfied by weekly, biweekly, monthly, seasonal, or even annual samples.
Summer phytoplankton communities can often replace themselves at 3- to 5-day
intervals and zooplankton populations may turnover at 5-day to 2-week intervals.
Summer periphyton communities can require two to five weeks to colonize and establish
reasonably stable communities on artificial substrates. Benthic invertebrates colonization
on such substrate can require similar periods, with their communities on natural
substrates undergoing nearly daily changes accompanying emergence of adults from the
water and migration by the remaining populations.
Defining locations to be sampled, frequency of sampling, and number of samples
to be collected depends on the area and range of conditions to be characterized through
monitoring. Further, when designing sampling networks, requirements inherent in the
statistical or other data analysis procedures intended to be used to evaluate the collected
data must be considered. For example, most simple statistical analyses are based on
samples collected at random times and locations encompassing the range of all possible
conditions of interest. Under these considerations, all possible locations and times
within the spectrum of possible combinations to be characterized should have equal
probabilities of being sampled. Thus, if the object is to characterize water quality
conditions "in the estuary," all locations in the estuary should, technically, have an equal
opportunity of being sampled to allow statistical inference of the sample results across
the area. (Here, a grid of randomized locations may be placed over a map of the
sampling area to identify sampling locations.) Similarly, if the objective to verify that
dissolved oxygen at point 1000 feet from a point source discharge remains at
concentrations greater than 6 mg/L, there ideally should be equal probabilities of
collecting samples any time over each 24-hr period, seven days a week.
Restricting the scope of possible sampling times and locations correspondingly
restricts the range of correct statistical inference. So, when samples are routinely
collected along the south shore from the designated monitoring location between 10:00
and 11:00 AM on the first and third Tuesday of each month, the collected samples
characterize only conditions during this one hour period on these two days per month
along the south shore at these locations.
5-18
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
In practice, most monitoring programs rarely adhere to these randomization
requirements strictly or even loosely. Typically, a set of "representative" monitoring
locations are defined, a "regular" schedule of sampling dates established, with sampling
beginning on each of these dates whenever the field crew can "get it together" in the
morning. Not uncommonly, "representative locations" are places having ready access.
Monitoring dates often occur early in the week to allow the remainder of the week for
analyzing collected samples. And, sampling crews general reach their first sampling
sites sometime between early- and mid lorning.
For such studies, conventional statistics remain appropriate to summarize and
characterize conditions at the locations and the times sampled. They also can be used to
evaluate possible differences among these locations and times sampled. It further remains
possible to produce deductive, subjective, and best-professional-judgement based
extrapolations to conditions at other times and places, but the claim cannot be made that
these qualitative extrapolations have a statistical, inductive basis. Finally, it should be
additionally noted that the strengths of deductive inferences can tend to increase when
similar results occur across increasing numbers of studies at the same sites or at other
sites.
Fortunately, there are various methods to increase the potential worth of collected
data and the ability extrapolate from these data with greater confidence by including
aspects of randomness within the monitoring design. For example, when the objective
of a study is reasonably served by regular spaced locations radiating out into the lake
or estuary from point source, the compass direction from the source for each successive
monitoring location might be selected using random procedures. Or, specific locations
for sampling collections may be positioned in "the representative reaches" using a
randomized grid method, with new sampling locations selected prior to each sampling
date.
To reduce sampling time bias in collected samples, the first of the predefined
monitoring locations to be sampled could be selected prior to each sampling trip using
randomized techniques. For example, say five routine stream monitoring stations are
identified as SI to S5, downstream. Always sampling SI first followed by sampling
from each subsequent downstream can result in an early sample bias (e.g., early morning
bias) at the upstream site, similar progressive time biases at each downstream site. But
the bias can be eliminated by randomly selecting the first of the five stations be sampled
and randomly determining whether the sampling occurs upstream or downstream.
Three sampling order randomization for three successive sampling trips might result in
sampling in the orders
S3, S4, S5, SI, S2;
5-19
-------
The Cadmus Group> Inc.
DRAFT
June 3,1993
• SI, S2, S3, S4, S5; and
S3, S2, SI, S5, S4.
Such efforts, however, can result in increased sampling costs associated with
travel times between stations, but these costs may not be significant, particularly if the
study design requires eliminating the sampling-time bias from collected data.
Sometimes intentional sampling-time bias can provide a useful feature in study
designs. For example, when low dissolved oxygen concentrations in receiving waters,
due to either organic or nutrient discharges, are a suspected problem requiring
assessment and monitoring, available data or professional judgement generally can point
to the area in the receiving water where the greatest oxygen depressions are most likely
to occur. Since lowest dissolved oxygen concentrations usually occur in surface waters
near and shortly following the dawn, sampling a station in that receiving water area
first, in the early morning, during each sampling trip can help to better understand the
occurrence, frequency and magnitude of the problem in the receiving water. The early
morning bias in the data from that station can be use as an indicator of "worse-case"
dissolved oxygen conditions over the extent of the receiving water body studied.
Overall, options for using randomized procedures are defined by the objectives
of the monitoring program, as limited by practical considerations regarding the nature
of the water body, possible access problems, and sampling costs. Thus, while
considerations and options for including randomization in monitoring programs can
provide valuable enhancements to the quality and die ability to extrapolate information
produced, including such procedures in these programs generally should not result in
sampling requirements that are either markedly more costly and less doable. Rarely,
and perhaps never, should demands for randomized field sampling procedures obstruct
completion of a monitoring program. The advice of Green (1979) is particularly relevant
here. "When possible emphasize use of statistics for choosing a sample number and
design that will allow convincing mean differences to be demonstrated without
additional statistics, rather than to prove slight, unconvincing differences."
Adding special, supplemental studies to assess conditions not included within the
routine monitoring program can also increase the potential confidence in .possible
extrapolations made using routinely collected data. For example, special studies can be
conducted at "key" sampling stations to evaluate diurnal trends in monitoring variables
collected at either random or regular intervals over a 24-hour period. (Here, random
selection of the sampling locations) and the first time of the regular sampling intervals
increases abilities to correctly use the results in statistical-based extrapolations of the
sampled data.) Other special studies can focus on following discrete slug flow down the
stream or through the reservoir or estuary.
5-20
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Determining the number of samples to be collected from each sampling
station on each sampling date to provide information on a monitored variable sometimes
is limited by practical considerations. For example, how much water can be reasonably
packed out of a wilderness area? Or, how much space exists on the specially
constructed artificial substrate holder for benthic invertebrate or periphyton samples?
A single sample provides only a point estimate about the state of the sampled variable
at the specific time and location sampled; no information is provided regrading the
possible variability that may exist in thi estimate. Compositing two or more samples
into a single sample provides information about "average" conditions from the sampled
locations when sampled, but the results still cannot provide quantitative information
about the variability inherent in the sample variable when sampled. Analyzing pairs of
collected samples again can provide information on "average" and a possible "range" of
conditions for the sampled locations, but again this sampling approach provides only
weak information on variability. Progressively adding more samples enables
determining progressively better information on variability.
The statistically based equations for determining sample size for a given
probability level require the specification of the equations in Box 5.1 and the following
quantities: WQ standard (C), the mean concentration where the standard should be
declared attained with a high probability (jit), the false positive rate (a), the false
negative rate (p), and the standard deviation ( G ). Box 5.2 gives an example of
calculating sample size.
5-21
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Box 5.1
Formulae for Calculating the Sample Size Needed to Estimate the Mean
nd = tf)
2 - + 2
1 - ft 1-a
5.1
where zH and zlKX are the critical values for the normal distribution with
probabilities of 1-a and l-£ (see any basic statistical text).
The sample size may also be written in the following equivalent form:
K - e J2 (Cs - ^)
ni
5.2
The term t (Greek letter tau) expresses the difference in units of standard
deviation.
5-22
-------
The Cadmus Group, Inc. DRAFT June 3,1993
Box 5.2
Example of Sample Size Calculations
Suppose it is desirable to verify compliance when the mean concentration is
.2 mg/L below the WQS of .5 mg/L (Cs = .5, ji, = .3) with a power of .80 (i.e.,
fJ = .20). Also suppose a = .43, a = .05, and 99 percent of the sample points will
result in analyzable samples/ then
x . (& - "¦) . . .465
a .43
From Standard Normal tables
z,_ = 1.645, z, , » 0.842.
2 %l 1-p
Using Equation 5.2 from Box 5.1,
' I* .465*
and
n *0±~1- 28.9.
f R .99
Rounding up, the sample size is 29.
Monitoring program designs should be clearly documentation. The importance
of this cannot be overemphasized. Without such documentation, changes in key
personnel can disrupt sample collection, resulting in failure to acquire needed
information or changes in collection or analyses methods leading to an additional source
of variation in the collected data. Also, during times of limited financial resources,
budget cuts can often fall on programs that lade dear, documented reasons for their
existence. Table 5-4 presents a checklist of topics under each of six general monitoring
components useful when developing and documenting monitoring program designs.
5-23
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
5.6 Special Waterbody Considerations
Receiving waters potentially requiring TMDLs include a diversity of
environmental types. Streams and rivers are linearly dynamic systems. Lakes are
primarily cyclic dynamic systems. Reservoirs and estuaries are a mix of both system
types, and are often influenced strongly by density gradients and currents. And
wetlands and riparian areas merge aquatic and terrestrial components. Some of the key
environmental characteristics for each of these types of environments, which must be
generally considered when designing monitoring programs and assessing impacts, are
briefly reviewed in this subsection. The key characteristics discussed are summarized
in Table 5-5.
5.6.1 Flowing waters - streams and rivers
Stream waters are intrinsically interconnected to their surrounding watersheds
(Motten and Hall, 1972; Hynes, 1975; Likens, 1984). Stream waters mostly originate as
runoff from overland flows or as groundwater seepage from the watershed. Hence, the
quality and quantity of water naturally reaching streams depend upon the infiltration
capacity of the soil and on other physical and chemical attributes of the watershed's
geology and soil lithology.
As water percolates over and through the soils, chemical constituents are
dissolved and other chemical reactions occur that define the natural water quality
characteristics in the stream. Within the stream, other mechanisms further affect the
composition of dissolved substances. For example, dissolved ions can be rapidly
adsorbed onto inorganic particles or absorbed by living organisms. These process also
the same properties that generally affect the abilities of flowing water systems to
assimilate waste chemicals and sediment discharged into them. Important considerations
involved in applying the TMDL framework to flowing water systems include stream
flow velocities, sediment transport, water temperatures, and material spiralling.
5-24
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Table 5.5. Key considerations by environment type important to the design of
water quality monitoring programs
Flowing waters - streams and rivers
o Watershed geology and aoil lithology
o Flow characteristics
o Sediment transport
o Water temperatures
o Nutrient and material spiralling
Lakes
o Water temperatures
o Thermal stratification
o Potential anaerobic processes in sediments and near bottom waters
o Seasonal vertical mixing of lake waters
Reservoirs
o Retention times for resident water
o Water temperatures
o Thermal stratification
o Intrareservoir water throughflow patterns
o Aging characteristics of reservoir
o Character of reservoir discharge structure
o Influences by sources/sinks characteristic on water qualities
o Downstream influences on thermal and chemical water qualities
Estuaries, harbors, and related near-shore brackish/marine waters
o Geomorphology characteristics, e.g, depth and mouth characteristics
o Upstream drift of marine waters
o Tidal waves
o Stratification and vertical mixing patterns
o Sediment transport and sediment-water interactions
o Mixing processes of fresh and sea waters
Wetlands and riparian areas
o "Nine" principal wetland functions
o Opportunities and effective in performing these functions
5-25
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Stream flow velocities may be thought of in terms of subcritical and supercritical
flows (Heede, 1980). Subcritical hows exert relatively low energies on banks and beds,
while supercritical flows can produce highly erosive forces and cause channel damage.
Standing waves are commonly associated with supercritical flows.
Sediment transport in streams is a very complex relationship involving at least 30
variables (Heede, 1980). But, in general, the mean particle size transported and sediment
mass discharged is proportional to the water volume discharged and the gradient of the
streamed. Also, water temperature can influence sediment transport For example, 40°F
water is able to carry two to three times more sediment than 80°F waters (Heede, 1980).
Flowing waters pose potentials for high erosion and production of deeply incised
channels and steepened valley slopes (Heede, 1980). To counter these potentials, natural
mechanisms exist that allow streams to adjust channel slopes, which helps to protect
streambeds. These mechanisms include (1) bed armoring by gravel and boulders, (2)
gravel bars that form transverse to stream flows, and (3) log steps that incorporate fallen
timber and associated debris into the streamed. Through such mechanisms streams can
reach a dynamic equilibrium with their channels (Heede, 1981).
Water temperatures in streams vary with air temperatures. This means that
stream temperatures have daily and seasonal cyclic patterns paralleling air temperatures,
but because of the greater density of water and the fact that water freezes at 0°C,
temperature extremes are less. Thus, in the northern temperate zone, stream
temperatures tend to be cooler than air temperatures during the summer and wanner
during the winter. Also, water temperatures tend to warm with distance downstream
in response to solar radiation and warm air; the rate of the warming is approximately
proportional to die distance traveled (Hynes, 1970). In stream reaches predominated by
groundwater contributions, stream temperatures tend to reflect the often cooler or
warmer groundwater temperatures.
Both cooling and warming trends can be important in triggering either
developmental or reproductive changes in aquatic organisms. Thus, daily and seasonal
cycles in stream temperature are frequently important in the development and growth
of aquatic invertebrates and fish.
In lakes and terrestrial ecosystems, nutrients move in cycles from organisms to
soils or sediment and then back to organisms again. Often these cycles can be
repeatedly completed within dose proximity of each other. But in rivers and streams,
the flow of water causes the cycles to be completed at progressive intervals downstream.
This downstream movement of these cycles, which include nutrient processing and
transport, is termed nutrient spiralling (Webster and Patten, 1979).
5-26
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
While moving downstream, materials can be transferred among environmental
compartments where nutrients can be "stored" for varying periods, effectively altering
rates for downstream movements. Storage times for different compartments can be
relatively long or short. For example, some nutrients incorporate mostly into the tissue
of organisms and are released bade into the water at relatively slow rates; while other
nutrients can be rapidly excreted by organisms, soon becoming available again for use
by other organisms. Such differences produce nutrient spirals of different spatial lengths
for different nutrients.
It has been suggested that shorter nutrient spirals help to establish constancy in
stream ecosystems, leading to potentially increased biomasses, spatial heterogeneity
through the stream continuum, resistance to external stresses, and ability to rapidly
recover from perturbations (O'Neill et al., 1979). Disturbances will tend to disrupt
storage mechanisms, increase losses of dissolved nutrients and/or nutrients in sediment,
and to increase spiral lengths within streams (Webster and Patten, 1979; Newbold et aL,
1983; Mulholland et al., 1985).
This valuable concept of material spiralling can also be applied to the transport of
other materials, including toxicants. As such, this recurring pattern of temporary storage
and repeated release of stored chemical should be carefully considered as a part of any
TMDL in flowing water systems.
5.6.2 Lakes
Thermal characteristics are, perhaps, the single most important physical influence
in lake environments. Water temperatures not only affect rates biological process at
which biological processes occur in lakes and other surfaces waters, but also heat
contents of the water also drive the primary processes mixing lake water and their
constituent chemicals.
Typically, temperate lakes are warmed by solar radiation during the spring. As
the cold (about 0°C) surface waters from winter are heated to 4°C (die temperature at
which water has its maximum density), these dense surface waters sink to die bottom,
pushing the colder, less dense bottom waters toward the surface. Following this spring
turnover of lake water, the lake first develops uniform temperatures extending through
the vertical profile of the water column. Then, as surface waters are gradually warmed,
lakes often become thermally stratified with a layer of warm water nearest the surface
(the epilimnvm), a middle layer where temperatures decrease at about 1°C par meter of
water depth (the metalimnion or thermocline), and a layer of cool water nearest the bottom
(the kypolimnion). Warming of surface waters tends to cause the metalimnion to sink
through the summer.
5-27
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
With the onset of autumnal cooling, the reverse of the above process occurs.
Surface waters cool until 4°C is again reached and these dense waters sink to the bottom,
displacing the warmer water upwards toward the surface where it cools. The fall
mixing of the lake leads again to uniform lake temperatures/ i.e., isothermal conditions.
Through the winter, as ice cover forms over the lake, lake waters often tend to
stagnate with relatively warm, dense (3-4°C) water on the bottom and cold, lighter (ca.
0°C) water near the ice-covered surface. During this winter stagnation period a chemical
stratification, of sorts, can develop, as will be discussed below.
Lakes that turnover twice yearly, such as just discussed, are termed dimictic lakes.
Lakes also exist that do not mix (amictic), mix once (monomictic), or multiple times
(polymictic). But dimictic lakes are the most common and are the ones most studied in
the north temperate zone. A comprehensive discussion of die thermal properties of
lakes is presented in Hutchinson (1957); good discussions are also found in most
limnology texts.
Most nutrients dissolved in the upper illuminated layers of a lake (the euphotic
zone) are incorporated into plant tissue through the process of photosynthesis by small,
free-floating algae (phytoplankton). Concurrently, many organic and inorganic pollutants
can be assimilated or absorbed into these growing cells or adsorbed onto them or
adjacent nonliving particles suspended in the lake water. Subsequently, many of these
algae are consumed by small floating animals (zooplankton) or fish, and many of die
zooplankton are consumed by larger animals, including fish.
Other phytoplankton and zooplankton die without being consumed Carcasses of
these dead organisms and fecal wastes from living organisms slowly sink toward die
lake's bottom. As these materials pass through the water column, chemical and bacterial
decomposition release some of the bound chemicals into die water column, where they
become readily available for reuse. The bulk of the settling material, however, passes
through the euphotic zone into the deep, often dark faphotic), waters of the hypolimnion.
Eventually, much of the material setdes in organic layers (detritus) on die bottom, where
it can become permanently stored.
Decomposition of detritus continues in the deep waters and on the bottom.
Because of low light intensities in these deep waters, photosynthesis usually cannot
occur. Without the continual input of oxygen through photosynthesis and with the
continual use of oxygen, through both decomposition of settled organic materials and
respiration by the resident organisms, concentrations of oxygen continually decrease in
the deep-water environment.
During periods of summer and winter stratification, oxygen can be totally
5-28
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
depleted in the deep waters. This depletion produces reducing conditions for chemicals
in the waters and in the sediment. Under such conditions, chemically bound oxygen
present in the settled organic materials can be used during bacterial metabolism. For
example, bacteria metabolize the oxygen contained in nitrate (NO3) and nitrite (N02)
compounds, which chemically reduces these substances into ammonia (NH3) and
ammonium (NH4) containing compounds (Wetzel, 1975). In this form the dissolved
nitrogen containing compounds readily move from sediment into the water column.
Similar processes can occur for other to:- : and non-toxic pollutants.
Phosphorus is present in aerobic sediment as both organic and inorganic
compounds, including apatite and orthophosphate ions covalently bonded to hydrated
iron oxides. In the reducing conditions produced by oxygen depletion, this organic
bound phosphorus is decomposed and phosphorus in ferric hydroxides and complexes
are reduced. Consequently, ferrous iron and absorbed phosphate are mobilized and
released into the overlying water. In contrast to die importance of bacteria in recycling
nitrogen compounds, they are relatively unimportant in the movement of phosphorus
compounds from sediment into the water column (Wetzel, 1975).
Once nutrients and other discharge pollutants are dissolved into the deeper layers
of the water column, the spring and fall turnovers can mix the redissolved nutrients up
through the entire column, where these chemicals again become available to affect the
biota inhabiting these upper water layers. Once exposed to oxygenated waters, however,
phosphates and some other metals rapidly react to form insoluble compounds and
complexes, in which form they again precipitate to the sediment Thus, the process of
recycling chemicals between die sediment and the biota in the upper water column is
repeated twice annually in dimictic lakes.
When applying the TMDL process to lakes, it critical to consider the seasonal
important of lake mixing, temperature, and chemical processed in the bottom sediment.
These processes are key to defining both the ultimate availability, fate, and effects of
many pollutants in lake ecosystems.
5.6.3 Reservoirs
Most simply, reservoirs are dammed rivers that form lakes. Consequently,
reservoirs include characteristics that are similar to those in both lakes and rivers. The
natural water quality in reservoirs, as in any surface-water body, is related to
(1) climate, especially the quantity and quality of incoming precipitation;
(2) the chemical nature of the geologic formations within the watershed basin
5-29
-------
The Cadmus Group, Inc.
DRAFT
June 3, 1993
through which waters drain before entering surface waters;
(3) vegetation within the basin and its influences, along with influences of
climate, on the structure and chemistry of basin soils; and
(4) influences of man within the watershed, i.e., contributions to surface runoff
produced by surface disturbances or effluents discharged directly into the
reservoir or its tributaries.
Construction and operation of reservoirs variously alter the natural quality of
water. Factors within reservoirs that can influence the extent of these alterations include
its shape (morphometry), retention time for waters within the reservoir, its age, thermal or
chemical stratification of the waters, biological activities, and discharge depth(s) and
timing.
A number of downstream hydrological changes are produced by all
impoundments, but the extent and timing of these changes vary among reservoirs:
(1) evaporation from reservoirs reduces the total average annual runoff from
watersheds;
(2) seasonal flows become less variable;
(3) annual extremes in flow are altered;
(4) magnitudes for floods attenuate; and
(5) pulses occur that are unnatural in terms of timing and duration (Petts,
1984).
Additional downstream effects often resulting from reservoir construction and
operation include alteration of
(1) daily and annual thermal patterns;
(2) nutrient, dissolved gas, suspended sediment, and salinity levels;
(3) composition of bottom sediment;
(4) shoreline stability; and
(5) composition of the biological community (Canter, 1985; Petts, 1984).
5*30
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
The multiple effects on the downstream biological communities include alterations
to the riparian vegetation, instream aquatic vegetation, composition and diversity of
resident organisms, migration routes for the stream's inhabitants, biotic productivity, life
cycles for the resident organisms, and trophic relationships among these organisms.
Dams also can cause permanent physical changes by increasing seismic
tendencies; altering groundwater flows and water tables; inundating settled areas;
destroying wildlife habitat; interfering with migrations of fish and other aquatic
organisms; increasing extensive aquatic weed growths; and potentially increasing the
spread of communicable diseases, particularly in tropical areas (Canter, 1985).
While thermal stratification is often the dominant physical force in lakes, through
reservoir water flow is often the dominant physical force within reservoirs. This is not
to say that thermal characteristics of reservoirs are not important. In fact, thermal
patterns in reservoirs can be very similar to those seen in natural lakes; and the flow of
water through reservoirs is often guided by thermal patterns within reservoirs. Also
similar to lakes, reservoir waters are primarily heated both by solar radiation and
through heat advection from inflowing waters; about 20% of die heat within a reservoir
can result from the latter mechanism (Whalen et al., 1982).
One approach to estimating whether a reservoir will stratify is to use the
densimetric Froude number (F). When F is less than 1/*, no stratification is expected
(Canter, 1985):
r _ 320 Z
-------
The Cadmus Group, Inc.
DRAFT
June 3, 1993
layers of deep-release reservoirs than occur in similar reservoirs or lakes having only
surface outlets. One consequence of these different patterns is the relatively more rapid
downward movement of the thermodine during the summer in those reservoirs with
deep-water outlets.
Perhaps the most significant aspect of summer and winter thermal stratification
in reservoirs is its influence on water flow through reservoirs (Neel, 1963; Wright, 1967;
Marcus, 1989). Upon entering a reservoir, tributary waters may flow through the
reservoir in a "river" near the surface as overflows, near the bottom as underflows, or
midway through the water column as interflows; or they may move down the reservoir
as discrete bodies of slowly moving water following prior inflows, while mingling
somewhat with waters both ahead and behind (Neel, 1963; Wunderlich, 1971).
Which path inflows follow through reservoirs depends on the relative densities
of both the inflow waters and the existing reservoir waters. These densities are, in turn,
mostly defined by their relative temperatures. But factors other than temperature,
including total dissolved solids (salinity), can also influence water density and flow
patterns (Wunderlich, 1971). Also other factors, in particular wind, can have
considerable periodic influences on the movement of water in lakes and reservoirs
(Hutchinson, 1957).
Cycles for nutrients in reservoirs can be quite similar to those in lakes, discussed
above. Oxygen contents in the hypolimnion of some productive reservoirs can be
depleted by the decomposition of organic materials settling from the epilimnion (Neel,
1963). Differences do exist, however, between reservoir and lake chemistries. First
construction and filling of reservoirs flood the former terrestrial environments.
Subsequent leaching of chemicals from the flooded soils and from rotting organic forest
debris can profoundly affect water quality in the reservoir. Decaying forest materials
consume dissolved oxygen and elevate carbon dioxide concentrations, while leaching can
extract dissolved nutrients and organic compounds from the flooded plants and soils.
As a result, heavy algae growths can be supported, undesirable levels of color and
odorous substances may be produced, and conditions that enhance aquatic productivity
or that can even be toxic to aquatic life may result (Sylvester, 1965; Canter, 1985).
Comparisons between areas in a reservoir where forest vegetation was left
standing to neighboring areas whore it was removed showed the former to be more
eutrophic than the latter (Hendricks and Silvey, 1977). These researchers also noted that
increases in the overlying water depths tended to decrease the influences of flooded
terrestrial materials on reservoir water quality. Over time, influences from the flooded
terrestrial environment on the overlying water quality decrease (Sylvester, 1965).
Nutrient regimes in reservoirs also can differ markedly from those found in lakes,
5-32
-------
The Cadmus Group, Inc.
DRAFT
June 3/1993
depending on the relative vertical location of the outlet through which reservoir waters
are discharged (Wright, 1967; Martin and Arneson, 1978). In effect, reservoirs with deep-
water outlets act as nutrient sources, while those with surface oudets act as nutrient
sinks, with respect to the downstream waters. Recalling the discussion for lakes,
nutrients contained in organisms and their wastes mineralize as they sink through the
water column, with the bulk of the mineralized nutrients becoming stored in the deeper
waters and sediment. Then with spring and fall mixing, these nutrient are recycled up
to water layers nearer the surface, and r ain become available for organic production.
In contrast, reservoirs having deep-water outlets continually discharge deeper waters
and the nutrients they contain (Wright, 1967). This continually removes nutrients from
the reservoir, preventing them from being recycled to producers in the upper water
layers.
Passage of water through reservoirs tends benefit downstream water quality by
reducing turbidity, hardness, and coliform bacteria levels; and by oxidizing organic
materials within the reservoir, potentially reduces downstream biochemical oxygen
demand (Canter, 1985). But reservoir passage also can be detrimental to water quality
by lowering reaeration rate for die water during storage, allowing buildup of inorganic
chemicals in the hypolimnion that can be released to enrich downstream waters, and
enhancing potentials for algae blooms (Baxter, 1977; Canter, 1985).
Perhaps the worst potential consequences of reservoir storage occur in the
hypolimnion during thermal stratification when dissolved oxygen decreases, anaerobic
waters develop, and iron, manganese and hydrogen sulfides can dissolve from the
bottom deposits (Neel, 1963; Canter, 1985). Also, by allowing surface discharges from
reservoirs to drop considerable distances into plunge pools, reservoir effluents can
develop supersaturated concentrations of dissolved gases, which can have substantial
adverse effects on resident stream biota (Petts, 1984).
5.6.4 Estuaries, harbors, and related near-shore brackish/marine waters
Estuaries are coastal water areas where fresh water, sea water, the atmosphere,
and sediment interact. Traditionally, estuaries are defined as semiendosed water bodies
that have a free connections to the open sea and, within which, sea waters are diluted
with fresh waters from land drainages (Fritchard, 1967). Classical estuary systems
include the lower reaches of rivers where saline and fresh waters meet and mix due to
tidal action. More recently this term has been extended to include additional marine
coastal waters, including saline bays and sounds receiving riverine discharges, and to
include backwaters of rivers discharging into the Great Lakes (Ambrose and Martin,
1990).
5-33
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Estuaries and related near-shore brackish/marine waters receive all pollutant
loads contained in their tributary inputs. Also, many of these waters are commonly
associated with municipal areas and ports, leading to additional direct loadings of
pollutants from adjacent point and nonpoint sources and from discharged shipping
wastes. Circulation patterns in estuaries often trap these nutrients, toxicants, and other
pollutants. Also, bottom sediment can store these pollutants, where they can be
transformed, again released to the water column, or buried. Other near-shore waters can
be affected by coastal currents.
The complex loading, circulation, and sedimentation processes occurring in these
waters present many difficulties for modeling water quality under waste load allocations
and the TMDL framework. Water transport and circulation processes in estuaries are
driven primarily by both river flow and tidal action. Upstream drift of heavier sea water
can produce longitudinal salinity gradients through the estuary. Where rivers flows are
strong and tidal mixing weak, relatively fresh waters can stratify over the denser saline
bottom layer. When this occurs, flow of bottom waters can transport saline waters and
pollutants upstream (much like the underflow pattern discussed above for reservoirs),
while surface layers transport fresh water and other pollutants downstream. In some
larger estuaries, coriolis acceleration, due to the rotation of the Earth, can significantly
deflect currents to the right in the northern hemisphere. Also, particularly in shallow
estuaries, wind can dominate estuarine mixing processes.
The geomorphology of estuaries strongly affects the transport of pollutants and,
consequently, their water qualities (Ambrose and Martin, 1990). For example, estuarine
depth controls propagation of tidal waves; shallow channels and sills increase vertical
mixing, while deeper channels promote stratification and greater upstream salinity
intrusions. Shallow sills near the mouth of estuaries can limit circulation and flushing
of bottom waters. Since each estuary is unique, the fundamental processes controlling
water quality in each must be evaluated individually. That is, to determine the fate and
effects of water quality constituents it is necessary to first determine processes
controlling their transport.
Often, one the most important aspect of water quality modeling in estuaries is (he
ability to successfully simulate sediment transport and sediment-water interactions, since
sediment can affect transparency and carry nutrients and toxicants into the water
column. Unlike fresh waters, which mai&tain reasonably constant water quality
conditions over extended periods, the frequent large changes in salinity and pH in an
estuary directly affect the transport behavior of many suspended solids. Many colloidal
particles agglomerate and settle in areas of significant salinity.
The mixing of fresh and sea waters in estuaries produces a dynamic continuous
variations in both space and time, with shifting patterns of potential impacts. For
5-34
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
example, nitrogen commonly limit algal and plant growth in sea waters whereas
phosphorus usually limits their productivities in fresh water; either nutrient may be
limiting in estuaries. Further, the interaction between nutrient limitations from these
sources may be additionally altered by nutrient from atmospheric inputs. Nearly 40%
of the nitrogen in Chesapeake Bay, for example, can come from wet and dry atmospheric
sources (Fisher et al., 1988).
Since pollutants entering estuaries can affect water quality both directly and
indirectly, evaluating potential effects of pollutant inputs should include both
quantitative and qualitative considerations. The principal concerns for water quality in
estuaries include:
• Salinity is an important determinate of available habitat and habitat
quality for estuarine organisms. High volume discharges can adversely
impact estuaries by decreasing salinities in marine estuaries or increasing
salinities in Great Lake estuaries.
• Sediment entering estuaries from various sources (e.g., tributary and
nonpoint source inflows) can alter benthic habitats and carry hydrophobic
organic chemicals, metals, and nutrients into an estuary. Also, upstream
movement of saline waters can carry contaminated sediment upstream.
Other transport process can mix pollutants from the bottom into the upper
water layers.
• Bacteria and viruses can enter estuaries with discharges from municipal,
industrial, and marina discharges and runoff from farms, feedlots, and
municipal areas. These pathogens are particular concerns for impact to
beneficial uses of beaches and other recreational areas, and of shellfisheries,
where accumulations of pathogens in shellfish can produce health threats
to humans.
• Dissolved oxygen depletion caused by oxidation of organic materials,
nitrification, digenesis of benthic sediment, and respiration excessive
growths of algae and higher aquatic plants can stress many aquatic
organisms. These problems can be aggravated by excessive loading inputs
of nutrients and organics from various sources, and waste heat discharges
from power plants.
• Nutrient enrichment, eutrophication, and overproduction caused by
excessive loading of nitrogen and phosphorus can produce problem
blooms of phytoplankton, excessive growths by other plant species, and
other problems discussed above, leading to disruption of the natural
5-35
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
communities.
• Aquatic toxicity by excessive ammonia, many organic chemicals, and metal
loads, often even at very low receiving water concentrations, can produce
widespread problems in estuaries that disrupt natural communities. Short-
term acute and long-term chronic toxicities can be affected by pH,
temperature, sediment concentrations, and various interactions among the
toxicants themselves.
• Bioaccumulation and effects to humans can occur as low concentrations
in waters are taken up and concentrated leading to concentrations in
organisms that are potentially toxic to fish predators, including humans.
Resuspension of these toxicants stored in bottom sediment can cause such
food chain effects to persist long after the time where problem causing
discharges are eliminated.
Because of these complex interaction among the possible mixing processes,
estuaries cannot be treated as simple advective systems, as is possible for most river
systems. Both advection and dispersion must be considered is selection appropriate
qualitative and quantitative models of estuarine processes (Ambrose and Martin, 1990).
5.6.5 Wetlands and riparian areas
In general, wetlands include lands where saturation by water is the primary
determinant affecting soil development and the composition of plant and animal
communities within the affected area (Cowardin et al., 1979). Water at least periodically
saturates or covers soils and substrates of most wetlands. This event severely stresses
all plants and animals not physiologically adapted to living in water or saturated soils.
Maintaining the physical, chemical, and biological integrity of wetlands, which are
counted among the Waters of the United States under the Clean Water Act, is often basic
to maintaining these qualities within their adjacent surface waters. Wetlands provide
important hydrologic functions in purifying waters and regulating water levels within
watersheds. For example, wetlands can remove and retain sediment and pollutants from
the water through the combined actions of sedimentation, biological assimilation, and
chemical decomposition (Mitsch and Gosselink, 1986). Sedimentation can occur as water
flow is restricted by dense vegetation. At these places much of the suspended load of
the stream is deposited on the wetland floor. Wetland plants can then take up
(biologically assimilate) the nutrients and some pollutants. This often occurs very
quickly because the high productivity of most wetlands allows many dissolved chemicals
to be rapidly incorporated into biomass. While nutrients and many pollutants can be
5-36
-------
The Cadmus Group, Inc.
DRAFT
June 3, 1993
released back into the water when the plant dies, much also is retained in wetlands
within peat accumulations. Chemical alterations of pollutants, including denitrification,
metabolic breakdown by microbes, chemical precipitation, and physical binding to
organic debris, are also important for pollution attenuation.
Mitigating effects of floods and storms in the watershed is another significant
hydrologic function of wetlands. Wetlands can mitigate flood effects by decreasing
water velocity and increasing the area /er which the water flows, thereby increasing
the rate of absorption of water into the soil As flood waters subside, wetland soils
release the stored water over a period of days to weeks, effectively "desynchronizing"
flood water flows. Maintaining riparian wetlands for flood mitigation is often less
expensive and more effective than building dams and other flood control projects.
Coastal wetlands, with their water-absorbing ability and wind-buffering vegetation, can
act as a barrier to storm surges that could otherwise flood the river basin farther
upstream.
Coastal and riparian wetlands are increasingly recognized as essential to the
survival of many commercial fish stocks because of their function as hatcheries. For
example, the precipitous decline of the Pacific Salmon populations are thought to be
partially due to the destruction of riparian habitats in the Pacific Northwest and northern
California.
As the number and size of wetlands decreases, their importance to endangered
flora and fauna increases dramatically. At present, wetlands are the primary habitat for
disproportionally large numbers of endangered and threatened species (Mitsch and
Gosselink,, 1986). This important function is increasingly making wetlands a
conservation imperative.
In all, wetlands have nine functions that potentially benefit water qualities and
beneficial used in receiving water ecosystems (Adamus et aL, 1991):
(1)
Ground water recharge,
(2)
Ground water discharge,
(3)
Flood flow alteration,
(4)
Sediment stabilization,
(5)
Sediment/toxicant reduction,
(6)
Nutrient removal and transformation,
5-37
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
(7) Production export,
(8) Aquatic diversity and abundance/ and
(9) Wildlife diversity and abundance.
While all wetlands tend to inherently possess all of these functions to some
degree, not all wetlands have equal opportunities to perform these functions nor are they
equally effective in their capability to perform these functions due to physical, chemical,
and biological differences in individual wetlands and in their surrounding environments.
These functions and their different opportunities and effectiveness that individual
wetlands have to provide these functions, in fact, form the basis for one common
technique for quantitatively evaluating wetlands (Adamus et al., 1987).
Overall, the major potential stressors that can affect the viability of wetlands are
increasing or decreasing water flows; dredging, filling, or other physical alterations; and
discharging nutrients or toxicants into the waters flowing into wetlands. Threats to the
condition and preservation of wetlands can come from any combination of these
stressors.
Riparian ecosystems occupy die interface surface-water and terrestrial ecosystems,
which contain distinct soil, vegetation, and, often, wildlife characteristics (cf., Mitsch and
Gosselink, 1986). They are associated with surface and subsurface drainage systems that
include perennial, intermittent, and ephemeral stream channels, ponds, lakes, reservoirs,
seeps, springs, and sinks, and are generally associated with high-water tables. They are
also characterized by structural and functional properties of, and interactions between,
both aquatic and terrestrial components. These systems are generally physically
bounded at their terrestrial edge by the uppermost floodplain or by die up gradient
extent of water available to vegetation through the lateral movement of groundwater
frdm the adjacent surface waters. Riparian areas are also bounded at the water's edge,
as defined by the low-flow surface water level, or the channel bottom in ephemeral and
intermittent streams. Their substrates characteristically have hydric or aquic properties,
and have the potential to support hydrophytic or phreatophytic vegetation. Energy,
nutrients, and species typically exchange continuously among riparian, aquatic, and
upland terrestrial ecosystems.
Riparian areas share the functional attributes listed above for wetlands. In fact,
palustrine wetlands, as defined by Cowardin et al. (1979) in the classification system for
wetlands used by the U.S. Fish and Wildlife Service, are a subset of riparian ecosystems.
When evaluating the assimilative capacities of receiving water systems, it is generally
valuable to evaluate the potential importance of wetlands and riparian ares in
contributing to this process. Ignoring the roles of these systems can lead to significant
5-38
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
underestimates of the potential assimilative capacities for the receiving water-wetland-
riparian systems. The methods presented by Adamus et al. (1987,, 1991) can be used in
completing assessments to evaluation these relationships.
5.7 Special Land-Use Considerations
Differing land uses also can provide unique challenges for monitoring plan
designs. Table 5-6 summarizes monitoring parameters of primary concern when
assessing impacts related to ten representative land uses. In addition to differences in
monitored parameters/ monitoring designs also can change to address different data
requirements. For example, monitoring programs for urban areas typically must address
concerns regarding potential individual and cumulative effects caused by multiple point
and nonpoint pollution sources. Similarly, generally a broad array of pollutants are of
potential concern for urban monitoring programs, including sediment, nutrients,
pesticides, metals, oil & grease, other toxicants, temperature, and BOD/COD
(Woodward-Clyde Consultants, 1990; U.S. EPA, 1993). Depending on the goals and
objectives of the monitoring program, individual monitoring stations may be located to
characterize:
(1) water quality and quantity contributions from each individual source;
(2) relative concentrations and changing characteristics along separate reaches
of the stream or river, or across individual section of the lake, reservoir, or
estuary;
(3) net changes in concentrations between upstream and downstream
sampling river stations, or between inflow and outflows for the lake,
reservoir, or estuary; or
(4) changing conditions through time at individual sampling stations within
any of die receiving water types.
Monitoring programs in agricultural area are generally concerned primarily with
investigating potential aquatic impacts from sediment, pesticides, and nutrients. These
pollutants enter aquatic ecosystems in these areas primarily through surface runoff,
although aerial application of either pesticides or, less often, nutrients can result in
airborne drift resulting in direct deposition of these chemical in surface waters.
Additional pollutants of concern in agricultural areas include organic (animal) wastes
and total salt loads (U.S. EPA, 1993).
5-39
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Table 5-6. Examples of monitoring parameters of primary concern to assess
impacts related to selected land uses (modified and expanded from
U.S. EPA, 1993)
Land Use
Chemical and Physical
Biological
Habitat
Cropland
Sadiment, nutrients, pesticides,
temperature
Dcnthlc macroin vertebrates,
ftlgM
Substrata composition, cover,
eutropiiiraHon
Grazing Land
Sadiment, nutrients,
temperature
MDuuc wmaamvwnmnam,
fiah, bacteria
Straambank stability, substrate
composition, cover, channel
chanderirttai
Feed lots
Sadiment, nutrients,
temperature, BOD/COD
Benthic macroinveftebrstsa,
bacteria
flnmtnnk ttaHHty, tubrtnt#
compoation, eutrophication
Urban areas
Sediment, nutrients, pesticides,
metals, oil * grease, other
toxicants, temperature,
BOD/COD
Benthk macro in vertebrates,
algae, Qah, bacteria
Streambank stability, substrate
compoethon, cover, channel
characteristics
Urban construction rites
Total suspended solids,
temperature
StraanriMnk stability, substrata
composition, cover, channel
characteristics
Highway*
Metala, trades, flow, sediment,
oil* grease
composition, oover, channel |
Forest harvMt
Sadiment, intergravel diaeohred
oxygen, temperature
Benthic macrotnvartebratas, fish
1tw.mh.nlr wlnf^f I
composition, cover, channel
characteristics
Forest road coiutruction and
maintenance
Sadiment, intergravel dissolved
oxygen, temperature
Fish, benthic
macroinvertabrataa, flsh
Channel characteristics,
aubatrata oompoattion,
Kreainbank stability, cover
Marinas
Metala, diaaotvad oxygen,
temperature, oil * grease,
BOD/COD, bacteria
Fecal coliform
Wetland vegetation, subatrsta
outnpotitioa, aover
Channelization
Flow, temperature, sadiment
Fiah, baathk
Aquatic vegetation, subatrate
cuuipueHkm, cow
In areas where the agricultural practices include provisions for irrigation return
flows, the qualities and quantities of a sampling of these discharge flows can be
monitoring to determine contributions to receiving waters. Where such practices are
lacking,, contributions to flowing water may be estimated through upstream-downstream
study designs to compute possible concentration differences and loading through the
5-40
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
reach of receiving water defined by the locations of the up- and downstream monitoring
stations. Non-point runoff to lakes, reservoirs, and estuaries are a more difficult
monitoring problem. Often the best that may be estimate the quality and quantity of
runoff by sampling receiving waters near representative location where runoff waters
enter the receiving water. Also, it also can be instructive to sample representative
surface-flow waters shortly prior to their entry into the receiving water, when such
surface waters are available for sampling. Total contributions can be roughly estimated
by extrapolation with total runoff voluir s calculated using precipitation volumes and
appropriate hydrologic models. Brakensiek et al. (1979) provide detailed descriptions of
methods and equipment needed for discharge monitoring use in both field and
watershed studies.
Timber harvest, silviculture, grazing, mining, oil and gas development, and road
and highway construction are the principal sources of impact to aquatic resources in
forests, rangelands, and other wildland areas. Potential aquatic contaminants of concern
include sediment, pesticides, animal wastes, and, generally to a lesser extent, petroleum
products. Often, the greatest concern regarding aquatic impacts in these wildland areas
involve physical impacts to the habitat and food resources of fisheries caused altered
flow patterns, breakdown of stable stream banks, loss of protective habitats (cover), and
excess deposition of fine sediment on stream bottoms causing loss of spawning habitat
and possible death or loss of benthic invertebrate food resources. MacDonald et al.
(1991) provide excellent monitoring guidance useful for assessing such impacts in forest
streams for the Pacific Northwest. Much of their guidance is useful in designing
monitoring programs to assess impacts to potential wildland aquatic resource from other
sources.
5-41
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
References. Chapter 5
Adamus, P.R., E.J. Clairain, Jr., D.R. Smith, and R.E. Young. 1987. Wetlands
evaluation technique (WET) - Volume 2: Technical Report Y-87. U.S. Army
Engineer Waterways Experiment Station, Vicksburg, MS.
Adamus, P.R., L.T. Stockwell, E.J. Clairain, Jr., M.E. Morrow, L.P. Rosas, and R.D.
Smith. 1991. Wetlands evaluation technique (WET) - Volume 1: Literature review
and evaluation rationale. U.S. Army Engineer Waterways Experiment Station,
Vicksburg, MS.
Ambrose, R.B., Jr., and J.L. Martin (eds.). 1990. Technical Guidance Manual for
Performing Waste Load Allocations, Book III: Estuaries, Part 1: Estuary and
Waste Load Allocations Models. EPA 823/R-92-002. Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
Armor, C.L., K.P. Burnham, and W.S. Platts. 1983. Field Methods and Statistical
Analyses for Monitoring Small Salmonid Streams. FWS/OBS-83/33. Office of
Biological Services, USDI Fish and Wildlife Service, Washington, DC.
Baxter, R.M. 1977. Environmental effects of dams and impoundments. Annual Review
of Ecology and Systematics 8:255-283.
Bordner, R., and J. Winter (editors). 1978. Microbial Methods for Monitoring the
Environment, Water and Wastes. EPA-600/8-78-017. Environmental Monitoring
and Support Laboratory, U.S. Environmental Protection Agency, Cincinnati, OH.
Brakensiek, D.L, H.B. Osbom, and W.J. Rawls. 1979. Field Manual for Research in
Agricultural Hydrology. Agricultural Handbook No. 224. Science and Education
Administration, U.S. Department of Agriculture, Beltsville, MD
Canter, L. 1985. Environmental impacts of water resources projects. Lewis Publishing,
Inc., Chelsa, MI.
Cowardin, L.M., V. Carter, F.C. Golet, and E.T. LaRoe- 1979. Classifications of
wetlands and deepwater habitats of the United States. U.S. Department of
Interior, Fish and Wildlife Service, Office of Biological Services, Washington, DC.
Driscoll, E.D., J.L. Mancini, and P. A. Mangarella. 1983. Technical Guidance Manual for
Performing Waste Load Allocations, Book n, Streams and Rivers, Chapter 2,
Biochemical Oxygen Demand/Dissolved Oxygen. EPA-440/4-84-020. Office of
5-42
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Water Regulations and Standards, U.S. Environmental Protection Agency,
Washington, DC.
Fisher, D., J. Ceraso, T. Mathew, and M. Oppenheimer. 1988. Polluted coastal waters:
The role of acid rain. Environmental Defense Fund, New York, NY.
Green, R.H. 1979. Sampling Design and Statistical Methods for Environmental
Biologists. John Wiley and Sons, New York, NY.
Hankin, D.G., and G.H. Reeves. 1988. Estimating total fish abundance and total habitat
area in small streams based on visual estimation methods. Can. J. Fish. Aquat.
Sci. 45:834-844.
Heede, B.H. 1980. Stream dynamics: An overview for land managers. General
Technical Report RM-72, Rocky Mountain Forest and Range Experiment Station,
Forest Service, United States Department of Agriculture, Fort Collins, CO.
Heede, B.H. 1981. Dynamics of selected mountain streams in the western United States.
Z. Geomorph. N. F. 25:17-32.
Hendricks, A.C. and J.K.G. Silvey. 1977. A biological and chemical comparison of
various areas of a reservoir. Water Research 11:429-438.
Hutchinson, G.E. 1957. A treatise on limnology. Volume 1, geography, physics, and
chemistry. John Wiley and Sons, Inc., New York, NY.
Hydroqual, Inc. 1986. Technical Guidance Manual for Performing Waste Load
Allocations, Book IV, Lakes and Impoundments, Chapter Toxic Substances
Impacts. EPA 440/4-87-002. Office of Water Regulations and Standards, U.S.
Environmental Protection Agency, Washington, DC.
Hynes, H.B.N. 1970. The ecology of running water. University of Toronto Press,
Toronto, Ontario, Canada.
Hynes, H.B.N. 1975. The stream and its valley Verhandlungen Internationale
Vereinigung fur Theoretische und Angewandte Limnologie 19:1-15.
Klemm, D.J., P.A. Lewis, F. Fulk, and J.M. Lazorchak. 1990. Macroinvertebrate Field
and Laboratory Methods for Evaluating the Biological Integrity of Surface Waters.
EPA/600/4-90-030. Environmental Monitoring Systems Laboratory, U.S.
Environmental Protection Agency, Cincinnati, OH.
5-43
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Likens, G.E. 1984. Beyond the shoreline: a watershed-ecosystem approach.
Verhandlungen Internationale Vereinigung fur Theoretische und Angewandte
Limnologie 22:1-22.
Martin, D.B. and R.D. Arneson. 1978. Comparative limnology of deep-discharge
reservoir and a surface-discharge lake on the Madison River, Montana.
Freshwater Biology 8:33-42.
Mills, W.B., G.L. Bowie, T.M. Grieg, K.M. Johnson, and R.C. Whittemore. 1986.
Handbook: Stream Sampling for Waste Load Allocation Applications.
EPA/625/6-86/013. Office of Water Regulations and Standards, U.S.
Environmental Protection Agency, Washington, DC.
MacDonald, L.H., A.W. Smart, and R.C. Wissmar. 1991. Monitoring Guidelines to
Evaluate Effects of Forestry Activities on Streams in the Pacific Northwest and
Alaska. EPA 910/9-91-001, Region 10, U.S. Environmental Protection Agency,
Seattle, WA.
Marcus, M.D. 1989. Limnological characteristics of a Rocky Mountain headwater
reservoir. Water Resources Bulletin 25(l):15-25.
Martin, J.L., R.B. Ambrose, and S.C. McCutcheon. 1990. Technical Guidance Manual for
Performing Waste Load Allocations, Book m, Estuaries. Part 2: Application of
Estuarine Waste Load Allocation Models. EPA-823-R-92-003. Office of Water
Regulations and Standards, U.S. Environmental Protection Agency, Washington,
DC.
Mason, W.T. (editor). 1978. Methods for the Assessment and Prediction of Mineral
Mining Impacts on Aquatic Communities: A Review and Analysis.
FWS/OBS-78/30. Office of Biological Services, USDI Fish and Wildlife Service,
Harpers Ferry, WV.
Mitsch, W.J., and J.G. Gosselink. 1986. Wetlands. Von Nostrand Reinhold, New York.
Motten, A.F. and C.A.S. Hall. 1972. Edaphic factors override a possible gradient of
ecological maturity indices in a small stream. Limnology and Oceanography
17:922-926.
Mulholland, P.J., J.D. Newbold and J.W. Elwood. 1985. Phosphorus spiralling in a
woodland stream: seasonal variations. Ecology 66:1012-1023.
CLAA
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Neel, J.K. 1963. Impacts of reservoirs. Pages 575-593 in D.G. Frey (editor). Limnology
in North America. University of Wisconsin Press, Madison, WI.
Newbold, J.D., J.W. Elwood, R.V. O'Neill and A.L. Sheldon. 1983. Phosphorus
dynamics in a woodland stream ecosystem: a study of nutrient spiralling.
Ecology 64:1249-1265.
O'Neill, R.V., J.W. Elwood and S.G. I lldebrand. 1979. Theoretical implications of
spatial heterogeneity in stream ecosystems. Pages 79-101 in G.S. Iimis and R.V.
O'Neill (editors). Systems analysis of ecosystems. International Cooperative
Publishing House, Fairland, MD.
Petts, G.E. 1984. Impounded rivers, perspectives for ecological management John
Wiley and Sons, New York, NY.
Plafkin, J.L., M.T. Barbour, K.D. Porter, S.K. Gross, and R.M. Hughes. 1989. Rapid
Bioassessment Protocols for Use in Streams and Rivers. Benthic
Macroinvertebrates and Fish. EPA/440/4-89/001. Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
Platts, W.S., W.F. Megahan, and G.W. Minshall. 1983. Methods for Evaluating Stream,
Riparian, and Biotic Conditions. General Technical Report INT-138,
Intermountain Forest and Range Experiment Station, USDA Forest Service,
Ogden, UT.
Platts, W.S. (and 12 others). 1987. Methods for Evaluating Riparian Habitats with
Applications to Management. General Technical Report INT-221. Intermountain
Research Station, USDA Forest Service, Ogden, UT.
Pritchard, D.W. 1967. What is an estuary: Physical Viewpoint In: GJi. Lauff, ed.,
Estuaries. Publication no. 83, American Association for the Advancement of
Science, Washington, DC.
Sylvester, R.O. 1965. The influence of reservoir soils on overlying water quality. Pages
327-335 in O. Jaag (editor). Advances in Water Pollution Research. Pergamon
Press, New York, NY.
Tetra Tech, Inc. 1992. Developing Nonpoint Source Load Allocations for TMDLs - A
Quick Reference Guide. EPA 841-B-92-001. Office of Wetlands, Oceans, and
Watersheds, Office of Water, U.S. Environmental Protection Agency, Washington,
DC.
5-45
-------
The Cadmus Group, Inc.
DRAFT
June 3, 2993
USDA Forest Service. 1989a. Fisheries Habitat Surveys Handbook. R4-FSH 2609.23,
Region 4-Intermountain Region, USDA Forest Service, Ogden, UT.
USDA Forest Service. 1989b. A Basin-Wide Stream Inventory Process Using Habitat
Type Classifications for Resident Salmonids. Medicine Bow National Forest,
USDA Forest Service, Laramie, WY.
U.S. Environmental Protection Agency (U.S. EPA). 1983a. Water Quality Standards
Handbook. Office of Water, U.S. Environmental Protection Agency, Washington,
DC.
U.S. Environmental Protection Agency (U.S. EPA). 1983b. Technical Support Manual:
Waterbody Surveys and Assessments for Conducting Use Attainability Analyses.
Office of Water, U.S. Environmental Protection Agency, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1983c Technical Guidance Manual
for Performing Waste Load Allocations, Book n, Streams and Rivers, Chapter 2,
Nutrient/ Eutrophication Impacts. EPA-440/4-84-021. Office of Water
Regulations and Standards, U.S. Environmental Protection Agency, Washington,
DC.
U.S. Environmental Protection Agency (U.S. EPA). 1984a. Technical Support Manual:
Waterbody Surveys and Assessments for Conducting Use Attainability Analyses,
Volume II: Estuarine Systems. Office of Water, U.S. Environmental Protection
Agency, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1984b. Technical Support Manual:
Waterbody Surveys and Assessments for Conducting Use Attainability Analyses/
Volume HI: Lake Systems. Office of Water, U.S. Environmental Protection
Agency, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1985. Guidance for State Water
Monitoring and WLA Programs. EPA 440/4-85-031. Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
U;S. Environmental Protection Agency (U.S. FPA). 1991a. Guidance for Water Quality-
Based Decisions: The TMDL Process. EPA-440/4-91-001. Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1991b. Technical Support Document
for Water Quality-Based Toxics Control EPA/505/2-90-001. Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
5-46
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
U.S. Environmental Protection Agency (U.S. EPA). 1991c. Watershed Monitoring and
Reporting for Section 319 National Monitoring Program Projects. Office of
Wetlands, Oceans, and Watersheds, Office of Water, U.S. Environmental
Protection Agency, Washington, DC.
U.S. Environmental Protection Agency. 1992a. NPDES Storm Water Sampling Guidance
Document. EPA 833-B-92-001. Office of Water, U.S. Environmental Protection
Agency, Washington, DC.
U.S. Environmental Protection Agency. 1993. Guidance Specifying Management
Measures for Sources of Nonpoint Pollution in Coastal Waters. EPA 840-B-92-002.
Office of Water, U.S. Environmental Protection Agency, Washington, DC.
Ward, R.C., J.C. Loftis, and G.B. McBride. 1990. Design of Water Quality Monitoring
Systems. Van Nostrand Reinhold, New York.
Warren-Hicks, W., B.R. Parkhurst, and S.S. Baker, Jr. 1989. Ecological Assessments of
Hazardous Waste-Sites: A Field and Laboratory Reference Document. EPA
600/3-89/013. Environmental Research Laboratory, U.S. Environmental Protection
Agency, Corvallis, OR.
Water Pollution Control Federation, 1983, Existing Sewer Evaluation and Rehabilitation,
Second Edition, Alexandria VA., Manual of Practice No. FD-6.
Webster, J.R. and B.C. Patten. 1979. Effects of watershed perturbations on stream
potassium and calcium dynamics. Ecological Monographs 49:51-72.
Wedepohl, R.E., D.R. Knauer, G.B. Wolbert, H. Olem, P.J. Garrison, and K. Kepford.
1990. Monitoring Lake and Reservoir Restoration. EPA 440/4-90-007. Prepared
by the North American Lake Management Society for Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
Wetzel, R.G. 1975. Limnology. W.B. Saunders Company, Philadelphia, PA.
Whalen S.C., S.A. Leathe, R.W. Gregory and J.C. Wright. 1982. Physico-chemical
limnology of the Tongue River Reservoir, Montana. Hydobiologia 89:161-176.
Woodward-Clyde Consultants. 1990. Urban Targeting and BMP Selection. Region V,
Water Division, U.S. Environmental Protection Agency, Chicago, IL.
5-47
-------
The Cadmus Group, Inc.
DRAFT
June 3,1993
Wright, J.C. 1967. Effect of impoundment on productivity, water chemistry, and heat
budgets of rivers. Pages 188-199 in Reservoir fishery resources symposium.
American Fisheries Society, Bethesda, MD.
Wunderlich, W.O. 1971. The dynamics of density-stratified reservoirs. Pages 219-231
in G.E. Hall (editor). Reservoir Fisheries and Limnology. Special Publication
Number 8, American Fisheries Society, Bethesda, MD.
5-48
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Chapter VI. Sampling and Data Collection
Purpose: This chapter discusses in greater detail the types of data likely to be
required for TMDLs, and methods of collection. Both the assembly of existing data and
sampling methods for the collection of new data are summamed.
6.1 Data Requirements
The guidance document for the TMDL process (U.S. EPA, 1991a) lists five steps
for a water quality-based approach to water resource protection. The first three of those
steps apply directly to TMDLs:
1. Identification of water quality-limited waters
• Review water quality standards
• Evaluate monitoring data
• Determine if adequate controls are in place
2. Priority ranking and targeting
• Integrate priority ranking with other water quality planning ad
management activities
• Use priority ranking to target waterbodies for TMDLs
3. Development of TMDLs
• Apply geographic approach where applicable
• Establish schedule for phased approach, if necessary
• Complete TMDL development
These three data steps are accomplished using a combination of existing data,
original data collected through monitoring programs, and modeling.
6.1.1 Identification of Water Quality-limited Waters
States need to examine the condition of water bodies and evaluate whether three
elements of water quality standards are supported:
• Designated uses (i.e. recreation, water supply)
• Chemical, physical, and biological criteria
• Anti-degradation requirement
6-1
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
Water quality data is collected and summarized under several programs. These
data sets will be useful when determining which water bodes are listed under Section
303.
Section 305 (b) Water quality assessment
Section 304 (1) Impaired waters
Section 319 Nonpoint source program
Section 314 Clean lakes program
Table 6-1 lists suggested screening analyses and data sources for determining
whether water bodies are impaired. Data sources are described in Section 6.2.
6-2
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 6-1 Screening Data for Listing of Impaired Waters
Impairment Category
Data Sources
Waters where fishing or shellfish bans
are in effect or are anticipated.
National Marine Fishery Service
(NMFS)
State Fish and Wildlife Agencies
State Environmental Agencies
State Health Departments
Local Health Departments
Waters where repeated fishkills or
physical abnormalities in aquatic life
have been observed during the last ten
years.
National Marine Fishery Service
(NMFS)
State Fish and Wildlife Agencies
State Environmental Agencies
State Health Departments
Local Health Departments
Recreational Fishing Groups
Commercial Fishing Groups
Waters where there are restrictions on
water sports or recreational contact
State Environmental Agencies
State and Local Health Departments
Local and Regional Park Departments
Waters identified by the Slate in its
most recent section 305(b) report as
either "partially achieving" or "not
achieving" designated uses.
305(b) reports
Waterbody System (WBS) Database
Waters listed under sections 304(1), 314,
and 319 of the CWA.
Waterbody System (WBS) Database
Waters listed by State as priority
waterbodies
State Water Quality Management Plans
Waters where ambient data, dilution
analyses, or effluent toxicity tests
indicate potential or actual exceedances
of WQ criteria
NPDES permit applications
Discharge monitoring reports
Pretreatment program data
Waters classified for uses that will not
support the fishable/swimmable goals
of lite CWA.
Section 305 (b) Water quality
assessment
6-3
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Impairment Category
Data Sources
Waters where ambient toxicity or
adverse water quality conditions have
been reported by local, State, EPA, or
other federal agencies, the private
sector, public interest groups, or
universities.
University researchers
NOAA
USGS
US Fish and Wildlife
NMFS
Waters listed as impaired by pollutants
from hazardous waste sites.
National Priority List prepared under
section 105(8)(A) of CERCLA.
Adapted from U5. EPA, 1991a.
Section 305 (b) Water quality assessment: States monitor the quality of their
waters and report the status biennially to the EPA. This information is compiled into
a biennial report to Congress. Information in the reports includes water-quality limited
water bodies, use nonattainment causes and sources, the magnitude of the cause, and
the source of the cause.
Section 304 (1) Impaired waters: Three lists of impaired waters (the mini, short,
and long) are submitted to the EPA by states. The mini list contains waters that the
state does not expect to achieve numeric water quality standards for priority pollutants
after technology-based controls are in place. The short list contains waters that are not
expected to meet standards because of toxic point sources, and the long list contains
waters that are not meeting the fishable/ swimmable goals of the CWA due to any
source of pollutants.
Section 319 Nonpoint source program: Under Section 319 states assess nonpoint
source pollution and report to the EPA. The report lists waters that will not meet WQS
without additional control action. The report also lists categories of nonpoint source
pollutants which contribute to the impairment of waters, procedures for implementing
BMPs, and control measures for reducing NFS.
Section 314 Clean lakes program: Under this program states were awarded
grants for studying publicly owned lakes. Although the extent of the studies vary
between states and lakes, many lakes were extensively studied under this program.
Water quality, biologic and morphological data were typically collected.
6-4
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
6.12 Priority Ranking and Targeting
Once impaired waters are listed under step 1, they are prioritized for further
action using a ranking procedure. The following ranking criteria are listed in Guidance
for Water Quality-Based Decisions: The TMDL Process (USEPA, 1991a):
• Risk to human health and aquatic life
• Degree of public interest a d support
• Recreational, economic, and aesthetic importance of a particular waterbody
• Vulnerability or fragility of a particular waterbody as an aquatic habitat
• Immediate programmatic needs such as wasteload allocations needed for
permits that are coming up for revisions or for new or expanding
discharges, or load allocations for needed BMPs.
• Water pollution problems identified during the development of the section
304(1) long list
• Court orders and decisions relating to water quality
• National policies and priorities such as those identified in EPA's Annual
operating Guidance.
A broad range of potential data sources will apply to this rather inclusive list of
ranking criteria. Studies conducted by federal and state agencies, universities, and
public interest groups should be examined.
6.1.3 Modeling Data Requirements for TMDLs
In the context of this section model is used to refer broadly to mathematical
representations of pollutant sources. These representations may vary from a simple
spreadsheet computation to a sophisticated computer model.
The data requirements of wet weather nonpoint source modeling efforts will vary
greatly depending upon what is simulated, which model is used, and the level of
accuracy required. Simple models require limited data collection and often rely on
assumed or user supplied transport coefficients or concentrations. For example, the
Federal Highway Administration (FHWA, 1990) predicts the pollutant load and
concentration washed from roadways to adjacent bodies of water. The model requires
measurements of right-of-way and pavement areas, the area of the watershed, general
traffic volume data, and basic rainfall recurrence data. The model predicts a probability
curve for pollutant washoff based upon either an assumed or user supplied event mean
concentration (EMC). This model requires relatively little data but produces a result that
is probably accurate only to an order of magnitude.
6-5
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
In contrast, a loading function model such as GWLF (Haith and Shoemaker, 1987)
requires a larger set of data. Daily precipitation and temperature data are required, as
are soil condition, land use, and agricultural practice data. The model has been
successfully applied to agricultural watersheds, and provides reasonable estimates of
monthly stream discharge and monthly export of nitrogen, phosphorus, and sediment.
More complex watershed scale models require much greater data collection efforts
and produce continuous results in contrast to the probability curve and the monthly
values produced by the two previous models. For example, DR3M-QUAL requires sub-
catchment areas, impervious area, channel slopes, lengths, and roughness, infiltration
parameters, channel dimensions, kinematic wave parameters, water quality parameters
including wash-off and build-up coefficients. The model will produce hydrographs and
pollutographs.
The Compendium of Watershed-scale Models for TMDL Development (USEPA, 1992c)
lists the data inputs required by 21 models useful for TMDLs. For specific data
requirements, consult model literature listed in the Compendium.
Table 6-2 lists meteorological and land data requirements used by many
watershed scale water quality models. Many of the data requirements can be met, in
many locations, by existing data. Typically, data collected in large scale data bases or
efforts will need to be cross-referenced with other available data and a portion of the
data should be field checked to assess its accuracy. As the spatially scale of data
requirements becomes finer (i.e., field scale versus sub-watershed scale), existing, broadly
collected data sets are typically less satisfactory and the need for field verification and
data collection increases.
Water and Sediment Quality Data: Useful water and sediment quality data are
shown in Table 6-3. These data may be available from prior studies, however, in many
cases, additional data collection will be required to adequately classify the water quality
of a water body. Most models will require a subset of this extensive list of parameters.
Reaction rates are used to tune a model to represent reality. They are typically arrived
at during the calibration process, although some of them such as dispersion can be
measured directly. Discussions of transformation rates useful in surface water modeling
are contained in Rates, Constants, and Kinetics Formulations in Surface Water Quality
Modeling (U.S. EPA, 1985). This manual also-contains numerous references for further
reading.1
1 In thu table, thnc typaa of bktchaaiicai oxygan damaad (BOD) in li*»d: total Wochaoucal oxygan damaad (BOD), cubooacaoua BOD (CBOD), tad ftttcgtnotu BCD
(NBOD). Cirbonaoaoua BOD a cauaad by hmratrapluc hactaria which art eapabb of uatag organic carton at an tmrgy nuna for nidation. NBOD it axattad by nitrifying
bacteria which oaddinfanm of ntosgan. la iwaata that contain* nitrogaii BOD ia initially the raauls of CBOD, but after aevwaldayfc BOD la laifaiy the mUt of the oxidation
of nitrogen form or NBOD. Thereaaon forth* lag to tha tttnt nquind by the nitrifying beaaria tar growth. Some models will only raquif* BOD data, while othaieuae CBOD
and NBOD to txamtM oxygw iynurta on i Owr nk. HmwiuttqiHi lor wnyttwf lor BCD, NBOD, wfl CBOD lit total, Iwwmr flu ltoniMiy«i«frtteMdintqmi
diffar.
6*6
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 6-2 General Data Requirements of Watershed Models
Topography
Watershed area
Sub-watershed areas
Channel lengths and slopes
Channel roughness (Manning coefficient)
Stream width and depth
Meteorological data
Temperature
Precipitation
Evaporation (ET)
Tidal height (for coastal systems)
Wind Data
Solar Insolation
Land Condition
Soil classifications
Soil conditions
Slopes
Vegetation cover
Hydrology
Low-flow data
Stage-discharge data
Mean flow data
High flow data
Land use practices
Area of land use by type
Impervious areas
Street cleaning practices
Agricultural practices
Population data
Housing data
Industrial sites
Waste Disposal Data
Wastewater point sources
POTW hydraulic capacities
Septic tank data
Industrial point and nonpoint source
data
Other significant nonpoint sources
6-7
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 6-3 Water quality data model requirements
Water Quality Data
Sediment Data
TDS
Porosity
Temperature
Grain size
TSS
Percent solids
Light extinction
Sediment oxygen demand
DO
BOD
Reaction Rates, Constants
COD
Physical dispersion
CBOD
Dissolved oxygen reaction rates
NBOD
Ph and alkalinity
Nutrients (P, N)
Nutrient dynamics
Chlorophyll
Algal dynamics
Periphyton
Coliform dynamics
Phytoplankton
Macrophyte growth
Alkalinity
Ph
TOC
Inorganic carbon
F.
ch
Bacteria
Metals
Organic Compounds
6-8
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
6.2 Assembling Existing Data
There are a wide variety of data sources available for examining watersheds,
quantifying pollutant loads, and examining water quality. For example, a single
watershed may have been sampled by university researchers, the USGS, and the State
environmental agency. Care should be taken when using existing data. Whenever
possible collection techniques, analytic techniques, hydrologic condition, and season
should be taken into account when usin and comparing data sets. Often monitoring
efforts several years apart will have used different collection or analytic techniques that
will render comparisons between the data meaningless.
6.Z1 General Availability of Data
Meteorological Data: Rainfall data are available from two federal sources.
The U.S. Department of Commerce National Weather Service operates thousands of
weather monitoring stations throughout the country. The National Climatic Data Center
(NCDC) provides weather data in computer-readable format and will supply a computer
program called SYNOP for analyzing and summarizing rainfall data. The NCDC can
be reached at (704) 259-0682.
Limitations of meteorological data: Precipitation and temperature values can vary
markedly over very short distances. As a result, regional weather data collected even
short distances from a watershed of interest can poorly represent actual precipitation and
temperature values. This can be tolerable for models such as GWLF which produce
monthly loading values, but can cause difficulty for modeling wet weather phenomena
such as CSOs and urban runoff.
Agricultural Data: Agricultural activities can result in significant nonpoint source
loads of sediment, nutrients, and pesticides. Land use data and cultivation practices are
required inputs for several watershed scale models that simulate die effects of these
activities. The regional Soil Conservation Service may keep records of land use for the
region. The local cooperative extension office will have information on general
agricultural practices for the region. The agriculture census published by the USDAis
also a general source of agricultural land use information. It provides information on
crops and livestock including land area planted and number of head, on a county basis.
The data is usually collected on a county basis, and must be converted to watershed
boundaries.
Hydrologic Data: The U.S. Geological Survey (USGS) maintains a nationwide
network of stream and river gauge stations that continuously monitor flow. Extensive
hydrologic data is available from regional offices. Statistics include high and low-flow
events, mean flow statistics, and flow event recurrence. Some gages can supply stream
6-9
-------
The Cadmus Group> Inc.
DRAFT
June 8, 1993
discharge values on a 15 minute interval. The USGS can also provide assistance in
transferring data from monitored to unmonitored watersheds. The USGS also collects
data on geology, water quality and land use.
The complexity of hydrologic data required will vary from model to model. For
mid-level watershed models, daily discharge values are sufficient to examine the
hydrologic performance of the model. Modeling of storm flows will require a much
finer time scale, since storm flows are short-term, dynamic events. Most dry weather
pollutant investigations use low flow event statistics such as the 7Q10 to examine peak
stresses to aquatic environments. The 7Q10 is the ten year recurring seven day low flow
value. Wet-weather investigations such as storm water and CSOs will use storm-related
flow statistics.
Other sources of hydrologic data include the SCS Hydrologic Bulletin, the U.S.
Agricultural Research Service, the U.S. Forest Service, and the U.S. Bureau of
Reclamation.
Soil Condition: The Soil Conservation Service publishes maps of soil types and
condition. These data are required by many watershed scale models to determine
erosion dynamics. Maps of soil classification are available for most locations. The SCS
supports local soil conservation districts. Some district offices also collect agricultural
practice information.
Topography: The USGS publishes 7.5 minute topographical maps for most
regions of the country, and has digital information files for use in GISs for many regions.
Population: The U.S. Census Bureau Data Publishes population and housing
data, by county and by census tract Census data can be linked to GIS (Geographic
Information System) software using TIGER software.
Land Use Data: Data from state, regional, and city planning agencies are useful
for developing land use estimates for watersheds. Land use data are also available from
the Land Use Data Analysis (LUDA) program of the USGS, the National Resources
Inventory of the US. Soil Conservation Service, the Bureau of the Census, and the
Census of Agriculture.
6.2.2 Databases
The majority of the following databases are national-scale data collection efforts.
Many of them contain valuable information, however, the quality and completeness of
the data varies, and they should be cross-referenced with local data sources to ensure
6-10
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
data correctness and completeness. Detailed references for the following databases are
contained in the Inventory of Exposure-Related Data Systems Sponsored by Federal Agencies
(U.S. EPA, May 1992d) and Office of Water Environmental and Program Information Systems
Compendium. (U.S. EPA, 1992e). While data extracted from these existing data sources
may not be sufficient to calibrate a complex water quality model, it is useful for simpler
screening models and may be useful both to scope and to supplement monitoring efforts.
There is significant overlap between several of the databases and the ability jto link
information by stream reach number.
Citv and County Files: These databases contain city and county information
including city and county names, latitude and longitude, census populations, and
Federal Information Processing Standards (FEPS) state and county numbers. Using these
databases, the user can link political boundaries (cities, counties) to river reaches. The
user can access the names of cities and counties by stream reach, and extract information
from STORET and Reach Files databases.
Contact. Bob King, Office of wetlands, oceans, and watersheds, Assessment and
watershed protection division, 202-260-7028.
Effluent GniHplinp* shiHips- This database is collected by the U.S. EPA Office of
Science and Technology to support industry specific technology-based effluent
guidelines. Effluent samples in the database are used to develop technology based
effluent limits for classes of industries and classes of wastewaters. The user can access
industrial sampling data by NPDES permit number, so the database may of use for some
location-specific studies. While this database will be of limited use to TMDL efforts, it
may provide supplemental data for screening efforts.
Contact: Eric Strassler, Office of science and technology, Engineering and analysis
division, 202-260-7120.
Gaye Files: This database contains hydrologic data for each reach in the EPA
Reach file. The data is collected at USGS stream gaging stations and modeled for
reaches where gages are not present Data available for lited stream reaches includes
gage station ID, type of data collected, collection frequency, mean and annual flow,
location, reach number, and 7Q10 low flow. The user can link to Reach files, STORET
data, and Reach file statistical and mapping tools (see separate descriptions).
Contact: Bob King, Office of wetlands, oceans, and watersheds, Assessment and
watershed protection division, 202-260-7028.
Lake Analysis Management System: This is a database of Great Lakes projects.
It contains physical and biological information on water, sediment, fish, and
6-11
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
phytoplankton from the Great Lakes Basin. The database contains data from 12 projects
including the Green Bay Mass Balance Project, and the Upper Great Lakes Channel
Study.
Contact Mr. William L. Richardson, U.S. EPA, 9311 Groh Road, Grosse lie, MI, 43138-
1697, 313-692-7611.
Reports: NTIS, 5285 Port Royal Road, Springfield, VA, 22161, 703-487-4650; Debra
Caudell, U.S. EPA, Environmental Research Laboratory, 313-697-7600.
Hvdrologicallv Linked Data Files (HLDF): This system links the following
databases:
• Reach File
• Stream Gaging Inventory (GAGE)
• Industrial Facility Discharge (IFD)
• Water Supply Data Base (WSDB)
Industrial Discharge File: This database tracks industrial point source
dischargers. There are approximately 120,000 records in the file. The database can be
cross-referenced to STORE! and Reach File data by either reach number or NPDES
permit number.
The user can obtain NPDES permit information and link to the Gage Hie, PCS and
STORET (see separate descriptions).
Contact Bob King, Office of wetlands, oceans, and watersheds, Assessment and
watershed protection division, 202-260-7028.
National Oceanic and Atmospheric Administration (NOAA): NOAA has several
environmental monitoring programs which store data on marine and estuarine waters.
The four databases and data compilations described below contain water quality, biotic,
and sediment quality data for major estuarine areas throughout the country. Several
pollution estimation efforts are part of the data compilation efforts. These estimates axe
probably most useful for screening waters for impacts and for determining which waters
to include in TMDL programs.
National Coastal Pollutant Discharge Inventory Program (NCPDD: This database
contains loading estimates for point, nonpoint, and riverine sources that discharge to
estuarine and coastal waters. Discharge loading estimates for nine classes of pollutants
are included in the database. At present the estimates represent the period 1980 -1085.
The source categories for the estimates include point sources, urban and non-urban
nonpoint sources, upstream sources, irrigation return waters, oil and gas operations.
6-12
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
marine transportation operations/ accidental spills and dredging operations. Sources of
land use data for the estimates include the Land Use Data Analysis (LUDA) program
of the USGS, the National Resources Inventory of the U.S. Soil Conservation Service, the
Bureau of the Census, and the Census of Agriculture. Data from state, regional, and city
planning agencies was also used.
Reports: Dan Farrow, SEA, ORCA, NOAA, 1305 East-West Highway, SSMC4 9th Floor,
Silver Spring, MD, 20910 (301) 713 - 30C
National Estuarine Inventory: The National Estuarine Inventory (NEI) is a
database which combines information from a variety of NOAA sources including the
NCPDI, the NS&T, and the National Shellfish Register. Data from the NEI is
periodically published as a series of data atlases. Data collected through 1990 is in the
process of being published. As part of the NEI estimates of pollutant loading are made
for estuaries. Point source information is extracted from PCS, nonpoint sources are
estimated using the Simulator for Watershed Resources- Rural Basins (SWRRB) and
Natural Resource Inventory (NRI) data, upstream sources are estimated using USGS
data.
Contact: Dan Farrow, Strategic Environmental Assessment Division (SEA), Ocean
Resource Conservation and Assessment (ORCA), NOAA, 1305 East-West Highway,
SSMC4 9th Floor, Silver Spring, MD, 20910 (301) 713 - 3000.
National Shellfish Register of Classified Estuarine Waters (Register): The Register
database contains a variety of information on estuaries classified for shellfishing.
The Registry was originally established to inventory the area and classification status of
shellfishing waters. Recently water quality data has also been included in the Registry.
The 1990 Register contains information on point and nonpoint pollution sources, shellfish
productivity data, reasons for classification changes, and discusses of relationships
between shellfish productivity, pollution sources, classifications, and public expenditures.
Contact Eric Slaughter, NOAA, ORCA, 1305 East-West Highway, SSMC4 9th Floor,
Silver Spring, MD, 20910 (301) 713 - 3000.
National Status and Trends for Marine Environmental Quality (NS&T): The
NS&T is a combination of several programs aimed at documenting chemical
concentrations reported in marine organisms and sediments. It consists of the Mussel
Watch Program, the Benthic Surveillance Project, Biological Effects Surveys and
Research, Historical Trends Assessments, Specimen Banking, Regional Assessments, and
a Quality Assurance Program.
6-13
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Contact Thomas O'Conner, NOAA, ORCA, 1305 East-West Highway, SSMC4 9th Floor,
Silver Spring, MD, 20910 (301) 713 - 3000.
National Water Quality Networks Program: This program includes the National
Stream Quality Accounting Network (NASQUAN) which is intended to detect trends in
water quality of surface waters and to relate those trends to upstream land and water
use. Data are stored in the USGS WATSTORE and EPA's STORET data bases (see
separate entries). Data can also be accessed through the USGS National Water Data
Exchange (NAWDEX).
Contacts: NAWDEX, 703-648-5664; Information guide, call USGS Chief Hydrologist for
Operations, 703-648-5031.
Ocean Data Evaluation System: The Office of Wetlands, Oceans, and Watersheds
(OWOW) maintains this database on marine water quality. It contains data from the
National Estuary Program, the Great Lakes National Program Office, the Ocean Disposal
Program, the 301(h) sewage discharge program, the NPDES program, and the 403(c)
(ocean discharge) program. Data is avaUable in electronic or hard copy format
Contact Bob King, Office of wetlands, oceans, and watersheds, Assessment and
watershed protection division, 202-260-7050; Tad Deshler, Tetra Tech, 206-822-99596. For
user ID contact Kim Stahlman 703-841-6005.
Permit Compliance System: PCS was created by the USEPA to meet the
informational needs of the National Pollution Discharge Elimination System (NPDES)
program under the Clean Water Act. The Office of Wastewater Enforcement and
Compliance (OWEC) maintains this database to track the permit, compliance, and
enforcement status of facilities regulated under the NPDES program. Only major
permittees (approximately 7,100 of 63,000 nationwide) are included in the database.
Regional USEPA offices and state users of PCS are responsible for the entry and
quality of the data in the system. The information contained in each record includes the
identity of the permitted pollutant, the discharge limits for that pollutant, the name,
address, and description of the discharging facility, and the measurements of the
permitted pollutant from discharge monitoring reports.
PCS contains only those chemicals which are regulated under the Clean Water Act
and permitted in waste water discharges. Inclusion in PCS indicates that the particular
chemical has been discharged into surface water, and is a possible source of drinking
water contamination. The pollutant measurements from the 1990 discharge monitoring
reports, given as concentrations, when multiplied by the flow, yield values for loading
to the nation's water.
6-14
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Contact: Dela Ng, OWEC (EN-338), U.S. EPA, 202-475-8323. For reports contact U.S.
EPA Library, Washington, DC—Freedom of Information Requests, or George Gray, 202-
475-8313.
Reach Files: This database contains information about segments or reaches of
waterbodies. The data include latitude and longitude of reaches, reach length, reaches
connected to the reach, and descriptions of water bodies. The Reach File system
includes mapping and analytic tools that allow linking to ARC/INFO based GISs.
Contact: Bob King, Office of wetlands, oceans, and watersheds> Assessment and
watershed protection division, 202-260-7028.
STORET: STORET is a database utility maintained by the USEPA for the STOrage
and RETrieval of parametric ground and surface-water quality data. STORET contains
water quality information for waterways throughout the country. The majority of the
data is from non-EPA sources including: the USGS, U.S. Army Corps of Engineers, U.S.
Forest Service, Bureau of Reclamation, and states. The data in STORET are stored in
several separate, but related systems. The systems pertinent for TMDLs are (1) the
Water Quality System, which records geographical, political, and descriptive information
concerning the sampling sites; and (2) the Biological System (BIOS) contains biological
information on aquatic organisms including distribution, abundance, condition, ad
habitat descriptions; (3) the Daily Flow System which contains daily flow information
from USGS gaging stations as well as some water quality information; and (4) the Fish
Kill File which contains information on fish kills resulting from pollution sources from
1960 through 1990.
STORET contains only those chemicals which are actively tested for in water; for
the most part chemicals which are already regulated or recognized as health and
environmental concerns. Appearance in STORET indicates that the particular chemical
was found in surface or ground water and is a possible contaminant source for drinking
water. Software links are being developed outside of the EPA to link STORET files to
GISs.
Contact: Thomas Pandolfi, U.S. EPA, Office of Water, Assessment, and Watershed
Protection, 202-260-7030. User registration Dora Craig, 202-260-7031.
Toxic Release Inventory (TRI): Section 313 of the Emergency Planning and
Community Right to Know Act, under SARA Title in, required EPA to establish a
computerized national database of toxic chemical emissions from manufacturing
facilities. This database includes information supplied by industries on releases of toxic
materials that occur during manufacturing and processing of materials. TRI is a
composite of more than 70,000 submissions of release reports filed on Section 313
6-15
-------
The Cadmus Group> Inc.
DRAFT
June 8,1993
chemicals. Release reports are required annually, and data are available through 1990.
Chemical release information includes quantitative estimates of the number of facilities
and the associated individual releases of chemicals to air, water, and land (data can be
separated on these three release medium categories). Facility information includes
location, industry by SIC code, storage quantity, and waste treatment data.
TRI is a voluntary reporting system for manufacturers and the quantities of
chemical released are estimated. The accuracy of these estimates probably varies greatly.
The number of facilities reporting may not be an exact indicator of the number of
releases because a facility may have multiple releases and companies with ten or fewer
employees are not required to report under this program. Since release quantities are
annual estimates, peak releases are not indicated. Inclusion in TRI shows that the
specific chemical has been released into surface or ground water or via land application.
The use of this database is probably limited to TMDL screening analyses.
Contact Steve Newburg-Rinn, U.S. EPA, Office of Toxic Substances, 202-382-3757.
Waterbodv System; The Waterbody System (U.S. EPA, 1991b) contains state water
quality assessment information collected to meet reporting requirements under section
305(b) of the Clean Water Act.
WATSTORE: The Water Data Storage and Retrieval System (WATSTORE) is the
USGS's repository for all of its water data including location, chemical and flow
information on surface and ground water. It is grouped into files:
Water Quality Hie
Groundwater Site Inventory File
Daily Values File
Peak Flow File
Water Use File
Station Header File
This database will be incorporated into the National Water Information System II which
is being developed by the USGS.
Contact: John Briggs, USGS, WRD, National Center, MS-437.12201 Sunrise Valley Drive,
Reston, VA, 22092, 703-648-5624.
6.23 Database Tools
The following database tools are used for linking data contained in several
environmental databases.
6-16
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Graphic Exposure Modeling System (GEMS); The GEMS system is an integrated
modeling tool for assessing surface and ground waters. It is linked to the EXAMS model
(a risk-exposure model) and to several dilution based surface water models including
PROUTE and REACHSCAN.
Contact: Sondra Hollister, U.S. EPA, Office of Prevention and Toxic Substances, 202-260-
3929.
Reach Pollution Assessment: This is a database tool that accesses and links data
from the Toxic Release Inventory (TRI), Permit Compliance System (PCS), the section
304(1) short list, STORET, the river Reach file, and the Industrial Facilities Discharge File.
The RPA can be used a screening tool to determine which pollutants are likely to be
found in a given area or river reach. It can help the user generate a map of water
quality sampling stations, industrial dischargers, POTWs, and released chemicals for a
given area. It does not however calculate or model the fate or transport of pollutants.
The RPA is accessed through the National Computer Center, through federal,
state, university and private communication networks. Users must be registered to
access the IBM 3090S computer at the NCC.
References: The Reach Pollutant Assessment User's Guide.
Contact Thomas Pandolfi, U.S. EPA, Office of Water, Assessment, and Watershed
Protection, 202-260-7030. User registration Dora Craig, 202-260-7031.
Environmental Display Manager: This display tool can be used to link STORET
and PCS databases, and can generate maps of permitted discharges, water bodies, and
population centers.
6.3 Sampling Techniques
This chapter provides general guidance on appropriate uses of alternative
sampling and data collection schemes. It reviews the mechanics and potential problems
of sampling. Selection and use of continuous monitoring devices, data recorders, and
data loggers is reviewed Special considerations regarding collection and handling of
chemical and biological samples are summarized with appropriate guidance to
additional sources for more detailed information. Conventional methods currently used
to assess habitat qualities and capacities are introduced. Unique safety concerns for
sampling during wet-weather conditions are reviewed.
6-17
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
6.3.1 Benefits and Limitations of Alternative Sampling Protocols
Monitoring samples are collected to gather physical, chemical or biological
information about a system. Most often, water or sediment samples or both are collected
from locations chosen during design of the sampling plan. They may be collected by
hand, manual samples; collected using by an automated sampler, automatic samples;
collected as individual point-in-time samples, grab or discrete samples; or combined with
other samples, composite samples. Samples collected and mixed in relation to the
measured volume within or flow through a system are commonly termed volume- or
flow-weighted composite samples, whereas equal volume samples collected at regular
vertical intervals through a portion or all of the water column may be mixed to provide
a water-column composite sample.
There are various purposes for collecting and analyzing samples. Concentrations
of contaminants in samples may be of interest for a variety of reasons including
comparison with a water quality standard or comparison with the known toxicant
tolerances for resident species. Mass loads of contaminants computed by multiplying
concentrations and flow rates may be used in assessing potential impacts from an
effluent discharge. The types, varieties, concentrations, and loadings to surface-waters
by contaminants of concern are needed when developing TMDL models and when
evaluating effectiveness and benefits from implementing treatment, control, and
management mitigation options.
Manual Grab Sampling, Samples collected by hand using various types of
containers or devices to collect water or sediment from a receiving water or discharge
are often termed manual grab samples. These samples can require little equipment and
allow recording miscellaneous additional field observations during each sampling visit.
Several advantages of manual sampling exist. They are generally uncomplicated
approaches, often inexpensive—particularly when labor is already available, and are
required for sampling some pollutants. For example, according to Standard Methods
(APHA, 1992), oil and grease, volatile compounds, and bacteria, must be analyzed from
samples collected using manual methods. (Oil, grease, and bacteria can adhere to hoses
and jars used in automated sampling equipment/ causing inaccurate results; volatile
compounds can vaporize during automated sampling procedures or be lost from weakly
sealed sample containers; and bacteria. populations can grow and community
compositions change during sample storage.) Disadvantages of grab sampling include
possible needs for personnel to be available around-the-clock to sample storm events and
possible exposure of personal to hazardous conditions during sampling. Also, long-term
sampling programs involving many sampling locations can be relatively more expensive.
6-18
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Grab sampling is often used to collect discrete samples that are not combined with
other samples. Grab samples also can be used to collect volume- or flow-weighted
composite samples, where several discrete samples are combined proportion to
measured volume or flow rates, however this often is more easily accomplished using
automated samplers and flow meters. Several examples of manual methods for flow
weighting are presented in the NPDES Storm Water Sampling Guidance Document (U.S.
EPA, 1992a). Grab sampling also may be used to composite vertical water-column or
aerial composite samples of water or sediment from various water kinds of bodies.
Automatic Sampling. Automated samplers have been improved greatly in the last
ten years and now possess features that are useful for many sampling purposes.
Generally, such sampling devices require larger initial capital investments or the
payment of rental fees, but they can reduce overall labor requirements costs (especially
for long running sampling programs), and increase the reliability of flow-weighted
compositing. Some automatic samplers include an upper part consisting of a
microprocessor based controller, a pump assembly and a filling mechanism, and a lower
part containing a set of glass or plastic sample containers and a well that can be filled
with ice to cool the collected samples. More expensive automatic samplers can include
refrigeration equipment in place of the ice well; such devices, however, require using a
120 volt power supply instead of a battery to operate. Also, many automatic samplers
can accept input signals from a flow meter to activate the sampler and initiate a flow
weighting compositing programs. Some samplers also can accept input from a rain gage
to activate a sampling program.
Most automatic sampler? allow either the collection of multiple discrete samples
or the collection of single or multiple composited samples. Also, samples can be split
between sample bottles or composited into a single bottle. Samples can be collected on
a predetermined time basis or in proportion to flow measurement signals sent to the
sampler.
When an automatic sampler is activated, the following sequence generally occurs.
The sampler pumps air through the collection line to displace any remains from a
previous sample and to clear debris from the end of the line. The sampler then draws
a sample according to the user defined program or flow meter signals. Finally, the
meter again pumps air through the line to expel any remaining water.
In spite of the obvious advantages of automated samplers, they include some
disadvantages and limitations. Some pollutants cannot be sampled by automated
equipment, unless only qualitative results are desired. While the cleaning sequence
provided by most such samplers provide reasonably separate samples, there is some
cross-contamination of the samples since water droplets usually remain in the tubing.
Often, debris in the sampled receiving water (e.g., plastic bags) can block the sampling
6-19
-------
The Cadmus Group> Inc.
DRAFT
June 8,1993
line and prevent sample collection. If the sampling line is located in the vicinity of a
flow meter, debris caught on the sampling line can also lead to erroneous flow
measurements. Thus, while automatic samplers can reduce manpower needs during
storm and runoff events, these devices must be checked for accuracy during these
events, and they must be regularly tested and serviced. If no field checks are made
during a storm event, data for the entire event may be lost. Thus, automatic samplers
do not eliminate the need for field personnel, but they can reduce these needs and can
produce flow weighted composite sample that may be either tedious or impossible using
manual methods.
Discrete versus Composite Sampling. Flow rates, physical conditions, and chemical
constituents in surface waters often vary continuously and simultaneously. This presents
a difficulty when attempting to determine water volumes, pollutant concentrations, and
masses of pollutants or their loads in the waste discharge flows and in receiving waters.
Using automatic or continuous recording flow meters allow obtaining reasonable and
continuous measurement of flow rates for these waters. Pollutant loads can be then be
computed by multiplying these flow volumes over the period of concern by the average
pollutant concentration determined from the discrete or flow composited samples. When
manual (i.e., instantaneous) flow measurements are used, actual volume flows over time
can only be estimated for loading calculations, but this approach adds additional
uncertainty into the loading estimates.
Where monitoring budget are unlimited, samples could be collected and analyzed
with a frequency approaching continuously. This would allow clearly establishing
patterns of changing chemical concentrations and physical conditions through time. But,
monitoring budgets are not unlimited—realistic sampling and analysis efforts must be
allocated to provide die maximum possible information for the minimum possible cost.
Analyzing constituents of concern in a single grab sample collection would
provide the minimum information at the minimum cost Such an approach could,
however, be appropriate where conditions are relatively unchanging, for example,
during periods without rainfall or other potential causes of significant runoff events.
Most often, the usual method for collecting samples would be to collect a random or
regular series of grab samples at predefined intervals during storm or runoff events.
When samples are collected often enough, such that concentration changes between
samples are minimized, then a dear pattern or time series for the pollutant's
concentration dynamics can be obtained. When, however, sampling intervals are spaced
too long apart in relation to changes in the pollutant concentration, less dear
understanding of these relationships are obtained.
Mixing samples from adjacent sampling events or regions (i.e., compositing)
allows analysis of fewer samples and, for some assessments, this is a reasonable
6-20
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
approach. Sample compositing provides a cost savings, especially related to costs for
water quality analyses, but it also results in loss of information. For example,
information on maximum and minimum concentrations during an runoff event is
generally lost. But, compositing many samples collected through multiple periods
during the events can help ensure that the samples analyzed do not include only
extreme conditions, not entirely representative of the event.
While analytical results from cc iposited samples will rarely equal "average"
conditions for the event, they still can be used, when a sufficient distribution of samples
are included, to provide "reasonably representative" conditions for computing loading
estimates. In some analyses, however, considerable error can be associated with using
analytical results from composited samples in completing loading analyses. For
example, when maximum pollutant concentrations accompany the maximum flow rates,
yet concentrations in high and low flows are treated equally, true loadings can be
underestimated. Consequently, when relationships between flow and pollutant
concentrations are unknown, it is often preferable to initially include in the monitoring
plan at least three discrete or multiple composite sample collections during the initial
period of increasing flow, during the period of the peak or plateau flow, and during the
period of declining flow.
The most useful method for sample compositing for flowing waters to assess
loadings is to combine samples in relation to the flow volume occurring during study
period intervals. There are two variations for accomplishing flow weighted compositing:
1) Collecting samples at equal time intervals at a volume proportional to the
flow rate (e.g., collect 100 ml of sample for every 100 gallons of flow that
passed during a 10 minute interval), and
2) Collecting equal volume samples at varying times proportional to the flow
(e.g., collect a 100 ml sample for each 100 gallons of flow, irrespective of
time).
The second method is preferable for sampling to assess loadings accompanying
wet-weather flows, since it results in samples being collected most often when the flow
rate is highest.
A third method is time-composited sampling, where equal sample volumes are
collected at equally spaced time intervals (e.g., collect 100 ml of sample every 10 minutes
during the monitored event). This approach provides information on the average
conditions at the sampling point during the sampling period. This approach should be
used, for example, to determine the average toxic concentrations that resident aquatic
6-ai
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
biota are exposed during the monitored event. Also, EPA recommends that to protect
against acute effects, the 1-hour average exposure should not exceed the critical
maximum concentration (CMC) more often than once every three years (U.S. EPA 1991c).
One-hour, time-composited samples will allow direct evaluation of this recommendation
for individual wet-weather events.
6.3.2 Precipitation Estimates
Rainfall data are available from two federal sources. The U.S. Department of
Commerce National Weather Service operates thousands of weather monitoring stations
throughout the country. The National Climatic Data Center (NCDC) provides weather
data in computer-readable format (U.S. EPA, 1992a) and will supply a computer
program called SYNOP for analyzing and summarizing rainfall data. The NCDC can
be contacted at (704) 259-0682.
It should be noted, however, that local rainfall can vary significantly from nearby
weather stations because the average storm cell is 4 to 5 miles in diameter, resulting in
large variations of intensity over short distances. Consequently, a long-term monitoring
plans often should include local rain gauges. Three types of recording rain gauges are
generally available: the tipping bucket gauge, the weighing gauge, and the float-
recording gauge. A thorough discussion of rainfall characteristics is presented in
Evaluation of Wet Weather Design Standards for Controlling Pollution from Combined Sewer
Overflows (U.S. EPA, 1992b).
6.33 Flow Rates and Receiving Water Volumes Estimates
Accurate flow monitoring is critical to compute dilution rates, mass loadings, and
predicting potential impacts in receiving waters. Selecting the most appropriate
monitoring technique depends on site characteristics, budgetary constraints, and
personnel availability. The following subsections introduce general options and
considerations useful when estimating flow rates and receiving water volume estimates
for receiving waters. The discussion conclude with list of several useful references
where more detail information can be obtained.
6.3.3.1 Existing Monitoring Stations
The U.S. Geological Survey (USGS) maintains a nationwide network of stream and
river gauge stations that continuously monitor flow. Various state and regional
organizations maintain similar networks of more limited geographical distribution.
6-22
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
When an existing gauge station is sufficiently near the points of interest in a stream or
river, additional flow monitoring in the stream may not be necessary. If the gauge
station is located in the same watershed, but up- or down-stream from the discharge, the
station's data can be converted to information relevant to the discharge point using an
appropriate transfer equation of the following form:
¦ *K)
Flow statistic desired for un-monitored watershed or sub-
watershed upstream of the CSO discharge.
Row statistic for a monitored regional, similar watershed/ or
sub-watershed.
Area of un-monitored watershed or sub-watershed.
Area of monitored watershed.
Regional flow exponent, derived for a regional network of
gauges.
A network of gauges can be similarly used to develop coefficients for the
following equation:
where C and p are regional coefficients developed by regression analysis.
Stream monitoring data collected in other regional watersheds can sometimes be
similarly transferred to other nearby watersheds when other information for the
watershed of interest is not available. It is important to remember, however, that
additional error is introduced into flow estimates when using these regional approaches.
For example, these equations are based on the assumption that the watersheds used to
develop regional coefficients have similar hydrologic responses. Local USGS offices and
USDA Soil Conservation Service (SCS) often can be consulted for help in appropriately
converting such data.
6.3.3.2 Flow Measuring Devices
When sufficiently information from an existing gauge station is not available, it
may be necessary to install new flow monitoring stations to address the needs of the
monitoring program. Flow monitoring requirements can vary from measuring flows
during individual sampling visits at each monitoring stations, to installing depth stage
where: Q«
Q«,
Aa
Ab
n
6-23
-------
The Cadmus Group> Inc.
DRAFT
June 8,1993
gauges (also called, "staff gauges") affixed to bridges or posts near the waters edge, to
installing complex and expensive stilling wells and continuous monitoring equipment,
or to a combination of several techniques.
Often it is useful to collect manual flow measurements using portable meters to
measure flow velocities. Although these methods can be time consuming, these devices
can provide valuable "snapshots" of the receiving water flows each time the monitoring
station is visited, can be used to characterize microhabitat conditions in the flowing
water that may be important to some monitoring objectives, and can be used to calibrate
stationary devices installed to monitor flow.
The primary stationary devices available include stage gauges, weirs, and flumes.
The main advantage of these devices is that the flow rate can be determined by reading
only the water depth, after water depth to flow volume relationships have been
developed. Flow measurements taken with these devices are significantly more accurate
than estimates based on flow-depth equations. Their primary disadvantage is that they
cannot measure flow where surcharging or backflow occur or where pooling occurs
immediately downstream of the gauging device. Also, the use of the last two devices
are primarily limited to moderate to small streams.
How (Velocity) Meters. Rotating, mechanical current meters, such as the Pygmy
or Gurley meter> arefrequently used to measure water velocity in streams and rivers.
Other more recently developed velocity meters use ultrasonic or electromagnetic
technology to sense flow velocity at a point, or in a cross section of a flow. Many of
these electronic devices are available as portable models or for permanent installation.
Velocity measurements must be combined with a depth values to compute flow
volume. Often the width of the stream is broken up into segments. Then the average
velocity in each segment of a stream or rivers is determined by measuring the flow
velocity at 0.2,0.4, and 0.8 of the maximum flow depth. The average stream velocity is
determined by multiplying each segment's velocity by each segment's area. These
products are summed and divided by the total cross-sectional area. Small shallow
streams are often measured by using a hand held gauge while wading and the velocity
of larger streams is measured from a bridge or a tethered boat. Extreme care should be
taken when working in or near moving waters.
Many of the newer meters combines velocity sensor with a depth sensors in a
single probe. For example, Ultrasonic Doppler sensors detect a phase shift between the
source and reflected signals, and electromagnetic meters measure the current generated
when the flow moves through a magnetic field. Velocity meters can measure flows in
a wider range of locations and flow regimes than can mechanical flow devices, and they
are not as prone to clogging as are the mechanical devices. They are, however, more
6-24
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
expensive equipment and can be inaccurate at low flows and when suspended-solid
loads vary rapidly. Installations using ultrasonic velocity sensors automatically switch
to a Manning Equation (see Section 6.3.3.3) when flow depth is too low for the sensors
to function properly.
Many flow meters can also be used to send signals to automated water quality
samplers, allowing the collection of water samples on a flow-weighted basis. Typically,
the user can calibrate these devices to >termine how much flow passes through the
flow meter, before die automated flow sampler collects draws the next water sample.
Velocity Measurement With Floats. Although less accurate than using current
meters, this technique is useful for rough flow estimates when current meters are not
available. Surface flow velocities can be estimated by using a slightly buoyant float,
such as an orange. While pieces of wood or grass at times also may be used, their use
should be limited and care must be employed to distinguish between movements of
these light objects by water versus wind. In use, the surface velocity is determined by
placing one or more floats in a stream and measuring the time needed to travel a
measured distance. Mean velocity for the stream can then be estimated by multiplying
the measures surface velocity by a coefficient, usually 0.85. Flow is the product of the
mean velocity and the cross-sectional area of the stream.
Stage-Discharge Gauging. Water flows are related to die stage, or the height of
the water in the stream channel: the higher the stage, the greater the flow volume. A
rating curve can be developed for each site where the stage (staff) gauge is installed by
making instantaneous stream flow measurements using portable methods and plotting
them against the stage of the stream at the time of measurement. A minimum of five
direct stream measurements of flow volume to stage should be made, with the
measurements describing the full range of stream flow conditions. Often, measurements
during the higher flows are die most difficult to obtain. If they are missing from the
stage-discharge plot, flow estimates during high flows may* be inaccurate because they
require extrapolations beyond the range of the data used to calibrate the stage gauge.
Once a stage-discharge relationship has been developed, flow rates can be
accurately estimated by reading the stage gauge periodic intervals or by recording water
depths using continuous recording equipment Automatic water-level recorders can
provide a continuous record of flow and do not require tending by field personnel
during storm events. They are, however, expensive and require regular maintenance
(see additional discussion below).
Weixs. Weirs are permanent or temporary devices placed across the flow channel.
They generally consist of a plate or wall structure with a notch, usually with a
rectangular or v-notched cut into their face. There are two main types of weirs—broad-
6-25
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
crested and sharp-crested. Broad crested weirs are dam like structures which pass flow
in a predictable manner. Masonry weirs can appear like brick walls and can behave in
a manner similar to broad-crested weirs. Sharp crested weirs can be fabricated from
sheet steel or plastic, and have a sharp top edge. Row rate can be determined by
measuring flow depth behind the weir, often by using a stage gauge or a continuous
depth (stage) recorded, and using the depth measurement in an appropriate formula,
such as the following:
Q = kHa
where: Q = Flow rate
k = Determined by unit conversions and weir dimension (usually
supplied by the manufacturer).
H = Height of flow surface above the bottom of the weir notch.
a = 1.5 for square notch weir, or 2.5 for v-notch weir.
Weirs can provide more accurate flow measurements than can flow-depth
equations. On the other hand, weirs cannot measure flow when they are flooded on the
downstream side, or when they are placed in pipes that are full, nearly full or flowing
backward. Weirs also are prone to fill with sediment or solids because flow velocity
drops to near zero upstream of the weir.
Flumes. Flumes are the most expensive flow gauging device. They are chute-like
structures inserted into a channel that force the flow into a known, uniform dimension,
allowing a better quantification of flow volumes. These structures include stage gauges
for manual calibration of depth-flow volume relationships. Also, the are often associated
with stilling wells to allow installation of continuous recorders for water depth/flow
volumes and, sometime, continuous sampling devices to assess water quality.
Flumes do not slow flows as much as do weirs and they are not as prone to
clogging and silting. Flumes can operate where some downstream flooding occurs,
when two depth measurements are collected. They do not work when completely
flooded or in backflow conditions. Two types of flumes are most commonly used, the
Parshall Flume and the Palmer-Bowlus Flume. The Palmer-Bowlus Flume is easier to
install than the Parshall Flume which requires that the floor of the existing pipe or
channel be dropped. For this reason the Paliner-Bowlus is commonly used for retrofit
and temporary applications.
6.33.3 Manual and Automatic Techniques
6-26
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Manual methods tend to cost less than automatic methods, and they allow
additional field observations during storm events. However, they rely more on using
well trained field personnel, particularly during storm events and they cannot provide
an accurate, continuous flow record. It is often useful to use manual flow measurement
as a screening tool when designing a monitoring plan. Monitoring plan designers can
use data generated by manual methods to allocate sampling equipment for maximum
benefit. Manual techniques are also valuable for checking the operation of automatic
flow meters during storm events. Personnel should be extremely careful when using
manual methods because conditions during storm events are often dangerous.
In general, manual methods are most useful for instantaneous flow measurement,
calibration of other flow measurements, and measurement of flow in small pipes. They
are of limited use in monitoring rapidly changing flows because an instantaneous
measure can be a poor indicator of the overall flow rate. Because wet weather events
occur during storms, use of manual devices can be hazardous to field personnel.
Automatic flow meters can efficiently collect continuous flow rate and volume
data, thus reducing the need for direct measurement by field personnel. Data can be
transmitted automatically over telephone lines, or accumulate on recording charts or
data loggers to be collected later by field personnel. Although automatic flow recorders
reduce the need for field personnel during storm events, field work, even during storm
events, is required to verify operation of the devices. In addition to retrieving flow data,
field personnel must also calibrate meters regularly. Debris can dog or alter meter
performance. Debris also can modify flow characteristics near the meter, rendering its
readings inaccurate. It is therefore important to operate redundant equipment, or
perform periodic manual measurements to verify the performance of automated meters.
In fact, when sampling resources are limited, using automated meters in critical areas
and manual techniques in less important locations may be cost-effective.
For recent information on continuous monitoring hardware, contact the USGS
Hydrologic Instrumentation Facility, Building 2101, Stennis Space Center, MS 39529.
Information on automatic gauging stations can be found in the U.S. Geological Survey
Water Supply Paper No. 2175 (Rantz, 1982).
Timed Flow. One of the most basic methods of flow measurement is the timing
of liquid flow into a container. This technique is accurate if performed carefully;
however, it is labor intensive and suitable only for small flows at discharge points. In
general, this method has limited use.
Dilution Method. A tracer, such as a fluorescent dye or salt solution, is injected
into the flowing water at a known rate. Field personnel measure die diluted tracer in
downstream samples using a fluoroscope (for fluorescent dye) or a conductivity meter
6-27
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
(for salt concentrations, although laboratory analysis gives more accurate results). The
flow rate is calculated from the measured dilution rate.
This method provides reasonable results for instantaneous measurements or
measurements during limited periods, but it is not appropriate for the continuous
monitoring of varying flows. There are additional drawbacks to this method. Dyes may
be adsorbed onto suspended solids or decay with time or because of sunlight, rendering
analysis results inaccurate. Industrial waste containing a high conductivity or salt load
can confound salt tracer results. Also, laboratory analysis of the salt concentration can
delay results of the flow measurement by several days.
Direct Measurement of Velocity and Depth. How rates (measured in volume
of flow per unit time) can be computed using a current meter to measure flow velocity
and a surveying rod to measure depth. Flow velocity (measured in flow per unit time)
varies considerably within a channel or pipe, with the slowest flow occurring next to the
banks, bottoms, and walls and the fastest flow occurring in the middle, as discussed in
Section 6.3.3.2. Depending on the width of the channel, one or more measurements must
be taken and adjusted to approximate the average cross-section velocity. Flow rate is
calculated by multiplying the computed average velocity by the cross sectional area.
While this type of measurement can be dangerous to perform during periods of high
flow, it can be useful for measuring flow in open channels and streams and for
calibrating other flow monitoring methods.
Flow-Depth Equations. Several equations, including the Manning Equation,
allow the user to calculate flow in an open channel or partially full pipe by measuring
the depth of flow. The basic parameters used in these equations include surface
roughness, diameter, and slope of the pipe. The Manning Equation is
V = JW*
6-28
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 6-4. Manning toughness coefficients for streams (modified from Dunne and
Leopold, 1978)
Streambed Characteristics Coefficient
Winding natural Stream and canals in poor condition, 0.035
considerable moss growth
Mountain stream with rocky beds; rivers with 0.040 - 0.050
variables sections and some vegetation along banks
Alluvial channels, sand bed, no vegetation
1. Lower reach 0.025
2. Upper reach
a) Plane bed 0.011 - 0.015
b) Standing wave 0.012-0.016
where: V * mean flow velocity (m/sec)
n = Manning roughness coefficient Based on type and condition
of channel, typically 0.02 to 0.05 (cf., Table 6-4)
R = hydraulic radius (m)
S = slope
(With V in feet/second and R in feet, the right side of this equation is multiplied by
1.486.)
Then, flow is computed as follows:
Q - VA
where: Q = volumetric flow rate
A = cross-sectional area
This technique can be used with either manual depth measurements or with
automated depth-sensing equipment The automated equipment can compile continuous
flow records using the Manning Equation.
6-29
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
The advantage of a flow equation is that it is simple to apply, requires that only
depth be measured, and needs limited equipment; however, the method is labor
intensive if performed manually. The main disadvantage of these equations is that they
rely on average characteristics of the channel network and, thus, are not very accurate.
Anomalies in slope, channel shape, or surface roughness often result in errors of 25
percent or greater.
Calibrating the equation to a particular location by using a second more accurate
form of flow metering can improve this method's accuracy. First a series of flow
velocities and depths are collected using a portable velocity meter and manual depth
measurement Instantaneous flow rates are computed by multiplying the velocities and
the depths and by using the flow-depth equation. If results from the flow-depth
equation do not agree with die direct measurements, a corrected equation can be
developed.
6.3.3.4 Rainfall-Runoff Flow Modeling
When field flow measurements are impractical or impossible, hydrological models
can be used to project stream flow volumes. These models require specific information
about the watershed upstream from section of channel of concern, including land use,
soil classification, slope, precipitation, and temperature. When these data are available,
watershed-scale models can do a reasonable job of predicting average stream flows. It
is more difficult, however, to predict periods of low and high flow without direct
streamflow information.
6.3.3.5 Dispersion Coefficient Measurement
As substances (for example, a pollutant plume) move downstream in a stream or
river their concentration decreases through two processes. It is diluted by inflow from
tributaries and groundwater, and it disperses or spreads in the direction of the river flow
(longitudinally). The longitudinal dispersion coefficient quantifies that spreading and
is a necessary input to many river models. The coefficient can be estimated by using
general river characteristics, or it can be determined using a dye study. The dispersion
coefficient can be estimated using the following equation:
E m SAxlQ-'l/1#
* HIT
where: Ex * Dispersion coefficient-miVday
6-30
-------
The Cadmus Group, Inc. DRAFT June 8,1993
U = mean river velocity-fps
B = mean width-feet
H = mean depth—feet
IT* = river sheer velocity-fps
and:
U* = JgHS
where: g = gravitation acceleration—32 ft/sec2
S = river bottom slope-ft/ft
The dispersion coefficient can be found by releasing a quantity of dye into a river.
The progress of the dye is monitored downstream with a fluorometer, and
concentrations and times of travel are recorded. The following equation is used to solve
for
M
s. ~
where: sp = peak dye concentration
M » mass of dye
A * cross-sectional area of the stream or liver
tp as time of travel for the dye peak
The peak concentration is plotted against the quantity 1A/L. The slope of the
plotted line is found from the line. The quantity E,is then solved for by the following
equations (Thomann and Meuller, 1987):
E - i(JL?
JOEl ' «
6.3.3.6 Flow Monitoring References
Considerable information on flow monitoring is available from additional sources.
A brief list of some of the available resources follows:
6-31
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
• ISCO. 1985. Open Channel Flow Measurement Handbook. Lincoln
Nebraska, (thorough discussion of the application of primary flow
monitoring devices; includes discharge charts for weirs and flumes)
• Parmley, R.O. 1992. Hydraulics Field Manual. McGraw-Hill, New York,
(contains useful hydraulic tables and charts; includes weirs, flumes, and
pipe flow)
• U.S. EPA. 1981. NPDES Compliance Flow Measurement Manual Report
No. MCD-77. Office of Water Enforcement and Permits Enforcement
Division, U.S. Environmental Protection Agency, Washington, DC.
(discusses the fundamentals of open channel flow monitoring)
• U.S. EPA. 1988. NPDES Compliance Inspection Manual. Contract No. 68-
01-6514,68-01-7050. Office of Water Enforcement and Permits Enforcement
Division, U.S. Environmental Protection Agency, Washington, DC.
(contains checklists for the correct application of flow monitoring
techniques and equipment)
• U.S. EPA. 1992. NPDES Storm Water Sampling Guidance Document.
Report No. EPA 833-B-92-001. Office of Water, U.S. Environmental
Protection Agency, Washington, DC. (discusses stormwater sampling,
monitoring techniques, and manual methods for flow compositing)
• Water Pollution Control Federation, 1983, Existing Sewer Evaluation and
Rehabilitation, Second Edition, Alexandria VA, Manual of Practice No.
FD-6. (discusses die fundamentals of open channel flow monitoring)
• Wedepohl, R.E., D.R. Knauer, G.B. Wolbert, H. Olem, P.J. Garrison, and K.
Kepford. 1990. Monitoring Lake and Reservoir Restoration. EPA 440/4-90-
007. Office of Water, U.S. Environmental Protection Agency, Washington,
DC. (contains useful descriptions of many of techniques for assessing
flows in of shallow streams and rivers)
6.3.4 Collection and Handling of Water Quality Samples
Samples are analyzed for the parameters identified in the monitoring plan. In
most cases, choosing an analytic method is a direct process. Samples collected for
compliance with NPDES programs, including TMDL, must be analyzed by a laboratory
certified by the appropriate authority, either the State or the U.S. EPA. The laboratories
must use analytic techniques listed in the Code of Federal Regulations (CFR), Title 40
6-32
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Part 136, "Guidelines Establishing Test Procedures for Analysis of Pollutants Under the
Clean Water Act" These methods are not, however, required for screening analyses to
determine merely the presence of a pollutant Also, a number of simplified methods
may be used to screen samples in the field. It is important to note that data generated
by non-approved methods may not be acceptable to EPA. Needs for EPA approval can
provide an important determinant in selecting methods for the collection, analysis, and
evaluation of water quality samples.
The balance of this subsection notes special considerations regarding those
parameters typically sampled and analyzed in the field, including pH, temperature, and
dissolved oxygen (DO). Samples collected for bacteria, cyanide, volatile organic
compounds (VOC), and oil and grease also require special collection techniques. For a
full description of collection techniques other sampling and analyses references should
be consulted (e.g., APHA, 1992; U.S. EPA, 1979a).
pH. Levels of pH can change rapidly in samples after collection. Consequently,
pH is often measured in the field using a hand-held pH electrode and meter. Electrodes
are easily damaged and contaminated, and must be calibrated with a standard solution
prior to each use. Held instruments also should be at thermal equilibrium with during
calibrations with the solutions being measured and when site measurements are
conducted.
Dissolved Oxygen. When multiple dissolved oxygen (DO) readings are required,
a dissolved oxygen electrode and meter (EPA method 360.1) is typically used. To obtain
accurate measurements, the Winkler titration method should be used to calibrate the
meter before and after each day's vise. If often is valuable to recheck the calibration
during the days of intensive use, particularly when the measurement are of critical
importance.
Oxygen electrodes are fragile, subject to contamination, and need frequent
maintenance. Membranes covering these probes should be replaced regularly, especially
when bubbles form under the membrane, and the electrode should be kept hill of fresh
electrolyte solution. If the meter has temperature and salinity compensation controls,
these should be used carefully according to the manufacturer's instructions.
Several manufacturers offer field kits to measure DO based on a modified Winkler
titration. In one method, the user collects a sample, adds three reagent packets, and
performs a simple titration. These kits are available for less than $50, while the cost of
a DO meter and its probe can range between $1,000 and $2,000.
Bacteria. The maximum recommended holding time for most bacterial tests is 24
hours, although most laboratories recommend that the sample be submitted for
6-33
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
laboratory analysis within 6- hours of collections. Bacteria samples are collected using
sterile containers or plastic bags/ and care must be taken not to let hands contaminate
the sample by touching the opening to the sample container or the sample as it enters
the container. Samples should generally be held on ice until processed.
Cyanide. Cyanide is reactive and unstable and must be carefully preserved.
Any residual chlorine or sulfide present must be removed when the sample is collected.
The sample pH is then adjusted to a pH greater than 12.0 using a caustic solution.
VOC Volatile organic compounds are collected in special glass vials that
facilitate analysis. Each vial must be filled so that there is no air space into which the
VOCs can volatilize and be lost. A sample is collected by filling both the vial and its lid
with water so that a meniscus forms at die water surface. The lid is placed on the vial
and screwed tight so no air space forms. If an air bubble is seen in the vial, the sample
must be recollected. VOC samples must not be composited in the field, but they can be
composited in a laboratory.
Oil and Grease. Oil and grease must be collected by grab sample using a glass
jar with a Teflon-coated lid. Samples are preserved by lowering the pH to below 2.0
using a strong acid.
6.3.5 Collection and Handling of Sediment Quality Samples
Receiving water sediments serve as sinks for a wide variety of materials.
Nonpoint source discharges typically include large quantities of suspended material that
then settle out in sections of receiving waters having low water velocities. Nutrients,
metals, and organic compounds can bind to suspended solids and settle to the bottom
of a water body when flow velocity is insufficient to keep them in suspension.
Contaminants bound to sediments may remain separated from the water column, or they
may be resuspended in the water column. Flood scouring, bioturbation (mixing by
biological organisms), desorption, and biological uptake all promote the release adsorbed
pollutants. Contaminants in sediments are especially available to enter organisms that
live and feed in sediment Having entered the food chain, contaminants can pass to
feeders at higher food (trophic) levels and can accumulate or concentrate in these
organisms. Humans also can ingest this contaminants by eating fish from waters with
contaminated sediments. Sediment depositions also can physically alter benthic (bottom
dwelling) habitats to affect habitat and reproductive potentials for many fish and
invertebrates. Sediment sampling should allow all these impact potentials to be
assessed.
6-34
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
6.3.5.1 Collection Techniques
Sediments samples are collected using hand or winch-operated dredges.
Although a wide variety of dredges are available, most operate in a similar fashion:
(1) The device is lowered through the water column by a hand-line or winch.
(2) The device is activated (e.g., released to allow closure) either by the
attached line or by a weighted messenger that is dropped down the line.
(3) The scoops or jaws of the device close either by weight or spring action.
(4) The device is retrieved to the surface.
Ideally the device disturbs the bottom as little as possible and closes fully so that
fine particles are not lost Common benthic sampling devices include the Ponar,
Eckman, Peterson, Orange-peel, and Van Veen dredges. When information about how
chemical depositions and accumulations have varied through time is needed, sediment
cores can be collected with a core sampling device. A thorough description of sediment
samplers is included in Macroinvertebrate Field and Laboratory Methods for Evaluating the
Biological Integrity of Surface Waters (Klemm, 1990).
Sediment sampling techniques are useful for two types of investigations related
to TMDL assessments: chemical analysis of sediments and investigation of benthic
macroinvertebrates communities. In either type of investigation, sediments from
reference stations should be sampled so that they can be compared with sediments in
the affected receiving waters. Sediments used for chemical analyses should be removed
from the dredge or core samples by scraping back the surface layers of the collected
sediment and extracting sediments from the central mass of the collected sample. This
helps to avoid possible contamination of the sample by die dredge. Sediment samples
for toxicological and chemical examination should be collected following method E1391
detailed in ASTM (1991). Sediments for population analyses may be returned in total
for cleaning and analysis, or may receive a preliminary cleaning in the field using a No.
30 sieve (see Section 6.3.7.4 and Klemm, 1990).
6.3.5.2 Sediment Analyses
There are a variety of sediment analysis techniques, each designed with inherent
assumptions about the behavior of sediments and sediment*bound contaminants. An
overview of developing techniques is presented in Sediment Quality and Aquatic Life
Assessment (Adams et al., 1992). EPA has evaluated 11 of the methods currently
6-35
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
available for assessing sediment quality (U.S. EPA, 1989). Some of the techniques may
be helpful demonstrating attainment of narrative requirements of some water quality
standards. Two of these common analyses are briefly introduced in the following
paragraphs.
Bulk sediment analysis tests for the presence of compounds in sediment material.
These tests typically analyze the total concentration of contaminant that are either bound
to sediments or present in pore water. Results are reported in milligrams or micrograms
per kilogram of sediment material. This type of testing often serves as a screening
analysis to classify dredged material. Results of bulk testing tend to over-estimate the
mass of contaminants that will be available for release or for biological uptake, because
a portion of the contaminants are not biologically available or likely to dissolve.
Elutriate testing estimates the amount of contaminants likely to be released from
sediments when mixed with water. In an elutriate test, sediment is mixed with water
and then agitated. The standard elutriate test for dredge material mixes 4 parts water
from the receiving water body with 1 part sediment (U.S. EPA, 1990). After vigorous
mixing, the sample is allowed to settle before the supernate is filtered and analyzed for
contaminants. This test was designed to estimate the amount of material likely to enter
the dissolved phase during dredging; however, it is also useful as a screening test for
determining whether further testing should be performed, and as a tool for comparing
sediments upstream and downstream of potential pollutant sources.
6.3.6 Water Quality Sample Preparation and Handling
Sample collection, preparation, preservation, and storage should minimize altering
sample constituents. Containers must be made of materials that will not interact with
pollutants in the sample and they should be cleaned in such a way that neither the
container nor the cleaning agents interfere with sample analysis. Sometimes, sample
constituents must be preserved before they degrade or transform prior to analysis. Also,
specified holding times for the sample must not be exceeded. Standard procedures for
collecting, preserving, and storing samples are presented in 40 CFR Part 136. Useful
material is also contained in the NPDES Storm Water Sampling Guidance Document (U.S.
EPA, 1992a).
Most commercial laboratories provide properly cleaned sampling containers with
appropriate preservatives. The laboratories also usually indicate the maximum allowed
holding periods for each analysis. Acceptable procedures for cleaning sample bottles,
preserving their contents, and analyzing for appropriate chemicals are detailed in various
methods manuals, including APHA (1992) and U.S. EPA (1979a). Water samplers,
sampling hoses, and sample storage bottles should always be made of materials
6-36
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
compatible with the goals of the study. For example, when heavy metals are the
concern, bottles should not have metal components that can contaminate the collected
water samples. Similarly, when organic contaminants are the concern, bottles and caps
should be made of materials not likely to leach into the sample.
6.3.6.1 Sample Preservation/ Handling, and Storage
Required sample preservation technique, and maximum holding times are
presented in 40 CFR Part 136. Cooling of samples to a temperature of £ 4°C is required
for most water quality variables. To accomplish this, samples are usually placed in a
cooler containing ice or an ice substitute. Many automated samplers have a well next
to the sample bottles to hold either ice or ice substitutes. Some more expensive
automated samplers have refrigeration equipment requiring a source of electricity. Other
preservation techniques include pH adjustment and chemical fixation. When needed,
pH adjustments are usually made using strong acids and bases. Extreme care should
be exercised when handing these substances.
Bacteria have a short holding time and are not collected by automated sampler.
Similarly, volatile compounds must be collected by grab sample as they are lost through
volatilization in automatic sampling equipment.
6.3.6.2 Sample Labeling
Samples should be labeled with waterproof labels. Enough information should
be recorded to ensure that each sample label is unique. The information recorded on
sample container labels should also be recorded in a sampling notebook kept by field
personnel The label typically includes the following information:
• Name of project,
• Location of monitoring,
• Specific sample location,
• Date and time of sample collection,
• Name or initials of sampler,
• Analysis to be performed,
6-37
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
• Sample ID number,
• Preservative used, and
• Type of sample (grab, composite).
6.3.6.3 Sample Packaging and Shipping
It is sometimes necessary to ship samples to the laboratory. Holding times should
be checked prior to shipment to ensure that they will not be exceeded. While waste
water samples generally are not considered hazardous, some samples, such as those with
extreme pH, will require special procedures. If the sample is shipped through a
common carrier or the U.S. Mail, it must comply with Department of Transportation
Hazardous Material Regulations (49 CFR Parts 171-177). Air shipment of samples
defined as hazardous may be covered by the requirements of the International Air
Transport Association.
Samples should be sealed in leak-proof bags and padded against breakage. Many
samples must be packed with an ice substitute to maintain a temperature of 4°C during
shipment. Plastic or metal recreational coolers make ideal shipping containers because
they protect and insulate the samples. Accompanying paperwork such as the chain-of-
custody documentation should be sealed in a waterproof bag in the shipping container.
6.3.6.4 Chain of Custody
Chain-of-custody forms document each change in possession of a sample, starting
at its collection and ending when it is analyzed. At each transfer of possession, both the
relinquisher and the receiver of the samples are required to sign and date the form. The
form and the procedure document possession of the samples and help prevent
tampering with the samples. The container holding samples can also be sealed with a
signed tape or seal to help ensure that samples are uncompromised.
Copies of the chain-of-custody form should be retained by the sampler and by the
laboratory. Contract laboratories often supply, chain-of-custody forms with sample
containers. The form is also useful for documenting which analyses will be performed
on the samples. These forms typically contain the following information:
• Name of project and sampling locations;
• Date and time that each sample is collected;
6*38
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
• Names of sampling personnel;
• Sample identification names and numbers;
• Types of sample containers;
• Analyses performed on each sample;
• Additional comments on each sample; and
• Names of all those transporting the samples.
6.3.7 Collection and Handling of Biological Samples
Biological assessment and toxicity testing are valuable tools for assessing toxic
impacts and may be required by some monitoring programs. This section introduces
biological sampling and laboratory toxicity testing.
Aquatic population evaluations used as a monitoring tool provide essential effects
information that no other evaluations may provide. This is because analysis of water
and sediment only provides concentration information for the analyzed sample(s), and
toxicity testing provides only effects information to particular species tested, which may
be more or less sensitive than the resident lake fauna. Resident populations and
communities of aquatic organisms, in effect, integrates over time all environmental
changes affecting these organisms. Thus, the biological community can reveal the
consequences of possible cumulative sources of impacts or short-term toxic discharges
not represented in the discrete collections of water samples.
Monitoring aquatic communities to assess effects generally requires that the
collected data be somehow direction comparable to previous data collected from the
same waterbody or from a similar waterbody, i.e., comparable to some reference
conditions (cf., Section 7.3). In fact, collecting data that is comparable between systems
is essential. Table 6-5 summarizes collection methods and types of information
potentially gained through monitoring. The balance of this subsection provides
additional introductory information on the four biological groups that are primarily
emphasized in monitoring programs. For additional detail beyond the information
provided here, the reference documents dted in Table 6-5 should be consulted.
6-39
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 6-5. Introduction to field sampling methods, and sources for more
detailed instructions on the proper use of each method
Sample
parameter
Information
Gained
Methods of collection
References
Phytoplankton
Algae
• Chlorophyll a
• Community
structure
• Primary
productivity
• Biomass
• Density
• Plankton buckets attached to a
vertical or horizontal tow net
(e.g. Wisconsin style net)
• Discreet depth samples using
VanDorn or Kemmer bottles
• Periphytometer
American Public Health
Assodation-APHA,
1992; American Society
for Testing and
Materials-ASTM, 1991;
Lind, 1985;
Vollenweider, 1969;
Weber, 1973;
Wetzel and Likens, 1979
Limitations:
Small organisms can pass through the net, and periphytometers are only good for
algae that attach to a substrate.
Riparian and
aquatic
macrophytes
• Community
structure
• Distributions,
depth & basin
wide
• Biomass
• Density
• Tissue
analysis
• Usually qualitative visual
assessments
• Quantitative assessments use
quadrant or line point
methods
American Public Health
Association--APHA
, 1992; American Society
for
Testing and Materials-
ASTM, 1991; Dennis
and Isam, 1984;
Vollenweider, 1969;
Weber, 1973; Wetzel and
Likens, 1979
Limitations:
Limited to the growing season for many species.
Zooplankton
• Community
structure
• Distributions
• Biomass
• Sensitivity
• Density
• Plankton buckets attached to a
vertical or horizontal tow net
(e.g. Wisconsin style net)
• Discreet depth samples using
VanDorn or Kemmer bottles
American Public Health
Assodation-APHA
1992; American Society
for
Testing and Materials-
ASTM, 1991; Und, 1985;
Pennak, 1989; Weber,
1973; Wetzel and Likens,
1979
Limitations:
Small organisms can pass through the net, some zooplankton migrate vertically in
the water column, therefore it is possible to miss some species.
6*40
-------
The Cadmus Group, Inc.
DRAFT
June 8f 1993
Table 6-5. (continued)
Sample
parameter
Information Gained
Methods of collection
References
Benthic
invertebrates
• Community structure
• Biomass
• Density
• Distribution
• Tissue analysis
• Ponar grab sampler
• Eckman dredge sampler
Surber
• Hess
• Kick net or D-ring net
• Artificial substrates
American Public
Health Association-
APHA, 1992; American
Society for Testing and
Materials—ASTM, 1991;
Lind, 1985;
Merritt and Cummins,
1984; Pennak, 1989;
Weber, 1973; Klemm et
al., 1990; Wetzel and
Likens, 1979
Limitations:
Some methods are time consuming and labor intensive, some methods are depth
restrictive (e.g. can only be used in shallow waters).
Fish
• Community structure
• Distributions, depth &
basin wide
• Biomass
• Density
• Bioconcentration
• Fecundity
• Hectroshocking
• Seines
• Gill nets
• Trawls
• Angling
• Traps
American Public Health
Assodation-APHA, 1992;
American Society for
Testing and Materials—
ASTM, 1991; Everhart et
al., 1975; Lagler, 1956;
Nielsen and Johnson,
1983; Schreck and Moyle,
1990; Ricker, 1975; Weber,
1973
Limitations:
Each method is biased to some degree as to the kind and size of fish collected.
Some methods are designed for use in relatively shallow water.
6.3.7.1 Fish
Of all the aquatic organisms/ fish generate the greatest public concern over
potential impacts. Consequently, fish can be the most important organisms to monitor.
Other groups of organisms may be more sensitive, but the effects of toxic chemicals on
fish are more readily seen. They include declines of populations and tumor growth on
individuals. For fish monitoring programs the principal characteristics of interest
include, (1) identification of species present, (2) relative and absolute numbers of
individuals of each species, (3) size distributions within species, (4) growth rates, (5)
6-41
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
reproduction or recruitment success, (6) incidence of disease, parasitism, and tumors,
(7) changes in behavior, (8) taste of fish, and (9) bioaccumulation of toxic constituents.
Common methods of sampling fish include angling, seines, gill and trap nets, and
electrofishing. The references shown in Table 6-5 provide guidance on methods used
for collection, measurement, preservation, and analyses of fish samples.
6.3.7.2 Phytoplankton
Phytoplankton, free-floating algae, are the principal primary producers, Le., the
primary source of photosynthesis, in most lakes and reservoirs. (Photosynthesis in some
shallow lakes is dominated by growths of algae on their bottoms or by larger aquatic
plants.) Phytoplankton are useful in monitoring and assessing aquatic conditions
because they include many species that are useful indicators of specific water quality
conditions, including the presence of some chemicals at concentrations not toxic to other
organisms. Phytoplankton also have relatively rapid rates of growth and population-
turnover (ca, 3- to 5-days during the summer season).
Laboratory analyses can provide information on the total numerical density
(number per water volume) of each phytoplankton taxa identified/ the relative
abundance by numbers or biomass of these taxa, the presence of or changes in indicator
species populations, and, less often, the total biomass of phytoplankton present. Lowe
(1974) and VanLandingham (1982) provide useful guides to the environmental
requirements and pollution tolerances of two key phytoplankton groups, respectively,
diatoms and blue-green algae.
6.3.7.3 Zooplankton
Zooplankton are microscopic animals that have only feeble swimming abilities.
As with phytoplankton, zooplankton are sensitive indicators of pollution and can
provide useful information about specific toxicants, particularly, in lakes and reservoirs.
Zooplankton are often collected by towing a plankton net though a measured or
estimated volume of water. Water volume can be measured by a flow meter set into the
mouth of the towed net. Water volume also £anbe estimated using the diameter of the
mouth of the net and the measured, or estimated, distance that the net is towed
vertically through the water column or horizontally behind a boat.
Laboratory analyses can provide information on taxa present, total numerical
density (number per water volume) of each taxon and their abundance by numbers or
6-42
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
reproduction or recruitment success, (6) incidence of disease, parasitism, and tumors,
(7) changes in behavior, (8) taste of fish, and (9) bioaccumulation of toxic constituents.
Common methods of sampling fish include angling, seines, gill and trap nets, and
electrofishing. The references shown in Table 6-5 provide guidance on methods used
for collection, measurement, preservation, and analyses of fish samples.
6.3.7.2 Phytoplankton
Phytoplankton, free-floating algae, are the principal primary producers, i.e., the
primary source of photosynthesis, in most lakes and reservoirs. (Photosynthesis in some
shallow lakes is dominated by growths of algae on their bottoms or by larger aquatic
plants.) Phytoplankton are useful in monitoring and assessing aquatic conditions
because they include many species that are useful indicators of specific water quality
conditions, including the presence of some chemicals at concentrations not toxic to other
organisms. Phytoplankton also have relatively rapid rates of growth and population-
turnover (ca, 3- to 5-days during the summer season).
Laboratory analyses can provide information on the total numerical density
(number per water volume) of each phytoplankton taxa identified, the relative
abundance by numbers or biomass of these taxa, the presence of or changes in indicator
species populations, and, less often, the total biomass of phytoplankton present. Lowe
(1974) and VanLandingham (1982) provide useful guides to the environmental
requirements and pollution tolerances of two key phytoplankton groups, respectively,
diatoms and blue-green algae.
6.3.7.3 Zooplankton
Zooplankton are microscopic animals that have only feeble swimming abilities.
As with phytoplankton, zooplankton are sensitive indicators of pollution and can
provide useful information about specific toxicants, particularly, in lakes and reservoirs.
Zooplankton are often collected by towing a plankton net though a measured or
estimated volume of water. Water volume can be measured by a flow meter set into the
mouth of the towed net. Water volume also £an be estimated using the diameter of the
mouth of the net and the measured, or estimated, distance that the net is towed
vertically through the water column or horizontally behind a boat
Laboratory analyses can provide information on taxa present, total numerical
density (number per water volume) of each taxon and their abundance by numbers or
6-42
-------
The Cadmus Group, Inc.
DRAFT
June 8, 1993
biomass, changes in major populations and indicator species populations, and, less often,
the total biomass of zooplankton present
6.3.7.4 Benthic Macroinvertebrates
Benthic macroinvertebrates, particularly aquatic insects, are an important food
source for fish and are widely recognized as useful indicators in aquatic monitoring
programs. These organisms can provide valuable information about the presence and
nature of toxics in the sediments of lakes and reservoirs. They live on and in sediments
where potentially toxic materials can accumulate and, like plankton, include many
important indicator species.
The benthic macroinvertebrate community is usually sampled using various
dredges, as discussed in Section 6.3.5. Samples are brought to the surface and either
preserved in their entirety in polyethylene bags or other suitable containers, or are
washed through a fine sieve (e.g., 500-pm screening or a No. 30 sieve) and then
preserved in a suitable container (Klemm, 1990).
Laboratory analysis can provide information on taxa present, the total numerical
density (numbers per sediment area) of each taxon, relative abundance by numbers or
biomass of these taxa, changes in major and indicator species populations, and the total
biomass of benthic macroinvertebrates present. A thorough discussion of the use of
macroinvertebrates in impact assessment is contained in Rapid Bioassessment Protocols for
Use in Streams and Rivers (Plafkii^, 1989) and Macroinvertebrate Field and laboratory Methods
for Evaluating the Biological Integrity of Surface Waters (Klemm, 1990).
6.3.8 Special Considerations for Collecting Wet Weather Data
6.3.8.1 Personnel Health And Safety
Hazardous conditions associated with sampling include:
• Hazardous weather conditions,
• Possible activity in confined spaces,
• Chemical hazards,
• Biological hazards,
6-43
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
• Receiving-water hazards,
• Traffic-related hazards, and
• Physical hazards.
Because the hazards listed above can be life threatening, safety should be of the
highest importance. The safety policies developed for a monitoring program should be
followed every time that personnel are in the field and emergency procedures should
be practiced ahead of time.
Some monitoring programs can require sampling in sewer or other enclosed waste
water discharges. Sampling such conditions can be hazardous even in dry conditions.
A sewer has limited access, little or no ventilation, little oxygen, and can contain toxic
or explosive gases. Rats and snakes may be present, and wastewater flows can contain
disease vectors and hazardous chemicals. Physical hazards are often present, including
high water flows, slippery surfaces, falling objects, and low overhead clearance.
Sampling storm-related flows presents hazards not encountered during dry
weather sampling. Visibility is often poor, and traffic hazards are increased, both in
driving to the site and working in street environments. Receiving waters, such as
streams and rivers, can be much higher than normal and present a drowning hazard.
Confined spaces. Confined spaces have poor ventilation and limited access.
Anytime personnel are below ground, a confined space plan should be in place and
followed. A written procedure should be developed and at a minimum, include the
following:
• Description of the type of work to be performed;
• Hazards that might be encountered;
• Location and description of the confined space;
• Information on atmospheric conditions in the confined space;
• Personnel training and emergency procedures; and
• Names of sampling personnel.
In general, the atmosphere of a confined space should be checked for oxygen and
hazardous gases every time that it is entered. Ventilation equipment should be used and/
6-44
-------
The Cadmus Group> Inc.
DRAFT
June 8,1993
in some cases, harnesses with lifting winches and self-contained breathing apparatus are
necessary. Personnel should be trained in confined entry procedures and should practice
confined space rescues at least once per year. In general, sewer systems should not be
entered during storm events.
When sampling in such spaces are required, the National Institute of Safety and
Health (NIOSH) manual, Working in Confined Spaces must be followed. The Occupational
Safety and Health Administration (OSH. is in the process of finalizing a confined space
entry permit system. This should be consulted, when it becomes available.
Traffic Hazards. Since sampling wet weather monitoring crews will be mobilized
during storm events, driving conditions will be hazardous and should be taken into
account Low-lying and coastal areas may present additional hazards, especially during
extreme storm events. If sampling will take place in traffic areas, warning signs and,
possibly, barricades should be used. When appropriate, a police officer or other
authorized traffic control personnel should be assigned to control traffic.
Chemical Hazards. Sampling personnel can be at risk from exposure to chemical
hazards in wastewater and in sample preservatives. Wastewater may be acidic or
caustic, or contain high concentrations of metals and organic compounds. Although
wastewater hazards are typically associated with industrial flows, hazardous household
wastes including herbicides and pesticides can pose a threat in areas without industrial
dischargers. Sampling personnel should wear long-sleeve clothing, gloves and safety
glasses. They should avoid skin contact with wastewater and preservatives. Personnel
should be trained in first-aid procedures for chemical burns and toxic exposure.
Biological Hazards. Biological hazards in sewers include vermin and disease
pathogens. Sampling personnel should be up-to-date with all appropriate vaccinations.
They should also avoid skin contact with waste water by wearing gloves and coveralls.
6.3.8.2 Placement of continuous monitoring equipment
Placing continuous monitoring equipment in the field prompt what can be aptly
described as a "extraordinary attraction for vandals and the preeminent destructive
forces of nature." It often seems that 100-year (greater) flood events have an
uhexplainable and uncanny nature of accompanying new monitoring programs.
Consequently and whenever possible, monitoring equipment should be installed in
inconspicuous locations, painted and otherwise treated to camouflage their appearance,
placed outside of 100-year flood channels, and anchored or attached (i.e., chained and
locked) to immovable objects (e.g., bridge pilings, larye trees, heavy anchors).
Reenforcing rod driven into and cut near the stream bottom can be used to help anchor
submerged monitoring devices that must be placed, for example, in the main channel
6-45
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
of streams. (Note: such rods should be removed at the completion of the study.) There
should be minimum use of cables or rods placed crosswise of water flows to minimize
trapping and accumulating floating debris. Use of deflectors on sampling equipments
and anchor devices also can help protect installed equipment from the effects of
destructive currents and debris. Often, long-term monitoring programs, depending on
the continuous collection of data for their success, should install permanent structures
made of concrete and steel to house and protect the monitoring equipment.
6.4 Toxicological and Habitat Evaluation
6.4.1 Toxicological Evaluation Methods
Aquatic toxicity testing has been used widely to assess the toxic effects of
chemicals and effluents on aquatic species. The role of toxicity testing in monitoring is,
most simply, to determine whether toxic chemicals are present in toxic concentrations.
A sufficient number of standardized tests are presently available to accommodate most
needs to assess potential problems from toxicants in aquatic environments. For example,
toxicity tests developed to evaluating chemically complex mixtures, such as effluents,
hazardous wastes, and sediments, are suitable for testing in TMDL studies.
In use, toxicity tests expose selected organisms to a range of concentrations for the
test chemical or effluent in solution. Generally, test concentrations bracket those
expected to occur in the receiving water. This range of concentrations is often prepared
by diluting the solution tested with either ambient receiving water or with standardized
laboratory dilution water. Effluent toxicity is assessed by comparing result obtained for
acute or chronic effects to test organisms in the various diluted concentrations of test
solution to results obtained in reference waters, which do not contain concentrations of
the test solution.
The preferred dilution water for the tests is ambient receiving water because it
contains chemical substances (e.g., suspended particles, dissolved organic carbon, and
various anions and cations) that may increase or decrease the toxicity and bioavailability
of test chemicals naturally in the environment. In contrast, use of standardized
laboratory dilution water may result in overestimates or underestimates of the toxicity
posed by the test chemicals in solution. Sometimes, contaminants or other water quality
conditions (e.g., very low ion concentrations) found in these waters may sometimes
prevent use of receiving waters. Then, use of uncontaminated surface waters from
neighboring lakes or streams that have otherwise similar chemistries may be appropriate.
Alternatively, the ambient dilution water's toxicity can sometimes be factored into the
assessment of the effluent's toxicity. This can be done by using toxic units to subtract
the toxicity of the dilution water from the toxicity of the effluent (U.S. EPA, 1991c).
6-46
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Acute and chronic tests are the primary methods use to determine aquatic
toxicities. Acute tests are "short-term" tests. When sufficient concentrations of toxic
chemicals are present, effects occur within a relatively short period of time (within 24
to 96 hours). Some chronic toxicity tests also can be relatively short-term tests (7-10
days, depending on the test species), or relatively lengthy tests as is the case for early
life stage tests or life cycle tests (30 days to one year or more, depending on the test
species). For acute tests, the endpoint is usually mortality, while endpoints for chronic
tests may include mortality, reduced reproduction, or reduced growth. Sources for
accepted standardized methods applicable to assessing toxic problems in lakes and
reservoirs include the methods developed for whole effluent toxicity testing by EPA
(Peltier and Weber, 1985, Weber et al., 1988, 1989) and the methods developed the
American Society for Testing and Materials for testing single chemicals, effluents, and
sediments (ASTM, 1991).
In general, toxicity tests to evaluate the potential toxicity of ambient water are
similar to effluent toxicity tests. Test organisms are placed in undiluted receiving water
for a specified period of time. Toxicity is then assessed by comparing die acute or
chronic effects observed in test organisms in the ambient water with effects observed in
organisms in water collected from reference sites or in laboratory control water. An
alternative method is to expose test organisms in situ in chambers placed in the receiving
water.
One key to successfully using toxicity tests to evaluate possibly contaminated
waters or sediments is by aptly defining the appropriate endpoints for the tests. For
example, selecting endpoints jhat measure, sublethal effects may tend to give "false
positives" because they are extremely sensitive to environmental perturbations. If test
endpoints are overly sensitive or not sensitive enough then a true prediction of toxicity
based on the test results will likely be incorrect. Some criteria for selecting measurement
endpoints for toxicity testing in both the laboratory and in situ are presented in Table
6-6. Common toxicity test species are listed in Table 6-7.
6-47
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 6-6. General criteria for measurement endpoints
• Readily measured
• Appropriate to the exposure pathway
• Appropriate temporal dynamics
• Low natural variability
• Diagnostic
• Broadly applicable
• Standard
• Existing data series
6.4.1.1 Good Laboratory Practices
Variability within result produced by standardized toxicity tests is inevitable.
Inherently, tests that involve living organisms and, often, complex mixtures of chemicals
that have some degree of variability. Explicitly defining specifications for completing
toxicity tests can help reduce the variability that may occur through differences in
applying procedures for each test method used. Although variability will never be
completely eliminated, consistent use of Good Laboratory Practices (GLP) by competent
personnel, will help to reduce variability.
GLFs are defined standards for laboratories. An integral part of GLPs is Quality
Assurance and Quality Control (QA/QC). QA/QC insures that the facilities, personnel,
equipment, methods, practices, records and controls conform to the rules and regulations
established by the US EPA for toxicity testing. Descriptions of QA/QC requirements for
aquatic toxicity testing are found in Peltier and Weber (1985) and Weber et al (1988,
1989).
6.4.1.2 Toxicity Identification Evaluations
While toxicity tests can identify whether toxicity is present, these procedures do
not identify the cause of the toxicity. Procedures developed by the U.S. EPA, called
Toxicity Identification Evaluations (TOss), can aid in identifying unknown toxicants(s). TIE
procedures combine short term toxicity tests with chemical and physical manipulations
of the solution, and analytical chemistry methods to identify toxics. These tests are
6-48
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 6-7. Common spedes used in aquatic toxicity testing
Common Name Scientific Classification
Vertebrates
Frog
Toad
Coho salmon
Rainbow trout
Brook trout
Goldfish
Fathead minnow
Channel catfish
Bluegill
Green sunfish
Invertebrates
Daphnids
Amphipods
Crayfish
Mayflies
Midges
Snails
Planaria
Ram sp.
Bufo sp.
Oncorhynchus kisutch
Oncorhynchus mysis
Salvelinus fontinalis
Carassius auratus
Pimephales promelas
Ictcdurus punctatus
Lepomis macrochirus
Lepomis cyanellus
Daphnia magna, D. pulex, D. ptdicaria,
Ceriodaphnia dubia
Gammarus lacustris, G. fasciatus, G.
pseudolimnaeus
Oronectes sp., Cambarus sp., Procambarus sp.,
Pasifasticus leniusculus
Hexagenia limbata, H. bilineata
Chxronomus sp.
Physa integra, P. heterostropha, Amnicola limosa
Dugesia tigrina
probably beyond the scope of most TMDL investigations; however, they may prove
useful in some studies to help separate impacts from individual sources from other
upstream sources of toxicity.
In application, the procedures involve up to three phases, characterization,
identification, and conformation. Phase 1 characterizes potential toxicants in the solution
6-49
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
through physical and chemical manipulations. These manipulations are designed to alter
or reduce the biological availability of chemicals such as oxidants, cationic metals,
volatile organics, non-polar organics, or metal chelates. Successful elimination of these
groups in each manipulation or fraction is subsequently tested by additional toxicity
tests. For more information refer to the Methods for Aquatic Toxicity Identification
Evaluations: Phase I Toxicity Characterization Procedures (Mount and Anderson-Carnahan,
1988).
The Phase n TIG procedures are designed to identify the suspected toxicants.
Separation and concentration steps are used to distinguish toxic from non-toxic
constituents. Separation methods may include simple filtration to more sophisticated
separation techniques such as High Performance Liquid Chromatography (HPLC), while
identification techniques may include Atomic Absorption Spectrophotometry (AAS) or
Gas Chromatography/Mass Spectrometry (GC/MS). Fractions are again tested in
toxicity tests to evaluate the success or failure of the methods. For more information
refer to the Methods for Aquatic Toxicity Identification Evaluations: Phase II Toxicity
Identification Procedures (Mount and Anderson-Carnahan, 1989).
The Phase III procedures are intended to confirm the suspected toxicant as the
true cause of toxicity. The final confirmation is needed not only to provide data to
prove that the suspected toxicants are the cause of toxicity in a series of samples, but
perhaps more importantly, to assure that the cause of toxicity is consistent from sample
to sample over time (Mount, 1989). Phase in incorporates several approaches to provide
a "weight of evidence" that the suspected cause of toxicity is indeed the actual cause of
toxicity, the approaches include correlation, observation of symptoms, relative
sensitivity, spiking, mass balance estimates and various water quality adjustments. For
more information refer to the Methods for Aquatic Toxicity Identification Evaluations: Phase
III Toxicity Confirmation Procedures (Mount, 1989).
6.4.1.3 Toxicity Reduction Evaluation
The Toxicity Reduction Evaluation (TRE) is a compliance mechanism that allows
dischargers to identify causes and develop corrective action for toxics as part of the
whole effluent toxicity (WET) requirements. TREs are designed to identify the causative
agents of effluent toxicity, isolate the sources of the toxicity^evaluate the effectiveness
of toxicity control options, and then confirm the reduction in effluent toxicity (U.S. EPA,
1991c). The TRE methodology, although designed for effluent toxicity, is applicable to
ambient receiving water toxicity and involves many of the steps to resolve toxic
contaminants in these waters.
6-50
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
6.4.1.4 Sediment Toxicity Testing
Sediment toxicity testing assesses the effects of sediment on the growth,
reproduction, and survival of aquatic organisms (Thomas et al., 1989). Many sediment
toxicity tests are similar to those used for ambient water toxicity tests methods, except
that for these the test sediments are often overlaid with water to assess the toxicity.
These tests are usually 10-day partial life cycle tests, which measure effects similar to
those presented above.
In bulk sediment toxicity testing, a sample of sediment is added to a known
quantity of water and sensitive organisms are added. When testing dredged materials,
a reference test of sediments collected outside of the dredging area is also performed.
Often, an upstream sample site can serve as a reference. Although the impact of a single
discharge is often difficult to assess with a high level of statistical certainty, using
upstream and downstream stations, a general comparison of the effect of the sediment
material on aquatic life can often be made.
The Draft Ecological Evaluation of Proposed Discharge of Dredged Material into Ocean
Waters (U.S. EFA„ 1990) presents a tiered approach to sediment testing when evaluating
ocean disposal of dredged materials. Some of the information also will be useful for
general sediment testing. The American Society for Testing and Materials procedures
for testing single chemicals, effluents, and sediments (ASTM, 1991) also provides useful
standardized methods. And, additional information on sediment testing and toxicity
assessment can be found in the text Sediment Toxicity Assessment (Burton, 1992).
6.4.1.5 Bioaccumulation
Bioaccumulation testing assesses whether a contaminant accumulates in the tissues
of an organism following exposure to collected water or sediment samples. Results from
similar analyses for samples exposure to waters or sediments from a reference site are
used for comparison. Test organisms are exposed to sample or reference conditions for
a specified length of time. After the exposure period, tissue samples are analyzed for
contaminants.
Fish and other organisms collected for bioaccumulation analyses either from
bioaccumulation studies or from the environment should be handled minimally and only
with clean hands or plastic gloves. The sampled fish should not touch metal boats,
buckets, or other containers that might contaminate the fish.
For fish specimens collected in the field, the species, length, weight, specimen
number, etc., of each sample should be recorded, the fish should be rinsed with water
6-51
-------
The Cadmus Group, Inc.
DRAFT
June 8f 1993
from the receiving water body, then placed individually into a clean polyethylene bag
until analysis. Each sample should be double bagged in a second bag, and immediately
stored on ice or frozen.
When analyzing metal accumulation in fish, fillets of muscle (meat) and skin,
specific organs (most often gills, liver, kidney), or bone may be used. Wearing clean
polyethylene gloves, laboratory technicians should take appropriate tissues samples from
each fish. Fish should be dissected on clean polyethylene work surfaces. Methods for
analyzing metal and organic contaminants in tissues are detailed by U.S. EPA (1991d).
6.4.2 Aquatic Habitat Assessment Procedures
The quality of aquatic habitats is defined through interactions among a diversity
of physical, chemical, and biological variables. Cultural developments have and will
continue to affect these natural interactions. Our knowledge of the natural range of
variation for these variables, how these variables interact, which of these variables are
most important, and how our cultural activities affect these natural relationships and
impact the quality of salmonid habitats is continuing to grow through extensive research
efforts and experience. Many questions remain.
It is dear, however, that successful survival and reproduction by aquatic
organisms is broadly defined by die physical structure of the environment, the quality
of the surrounding waters, and interactions with other organisms. The principal
characteristics of environmental structure that influence population abundance and
structure include riparian vegetation, channel morphology, streamflows, deposited
sediment, and winter snow and ice accumulation. Important water quality
characteristics include suspended sediment, temperature, pH, nutrients, and potentially
toxic chemicals. Biological influences involve nutrient and energy cycles, interactions
with invertebrates, competition with and predation by other fish, and predation by birds
and mammals. Additional, information to aid interpretation of many of these variables
individually is presented in Section 7.3.
Once receiving water conditions for these variables that constitute aquatic habitat
have been defined, the next step is to evaluate the'quality of the habitat based on some
measure of the variables. Following this, the variables can be monitored through time
to detect changes in habitat quality. Note however, that the definition of habitat quality
is difficult and equivocal, and that a single definition of quality does not directly apply
to all aquatic species or their habitats.
In practice, aquatic habitats are evaluated by first measuring physical, chemical,
and biological conditions in the study waterbody of concern. Often similar studies have
6-52
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
previously led to defining quantitative relationships (e.g., models based on regression
analysis) associating biological conditions in similar waterbodies (e.g., kinds of species,
numbers of species, and densities or biomasses of populations) to measures of physical
and chemical habitat conditions. When this information is available, physical and
chemical conditions measures in the study waterbody of concern can be compared to the
modeled conditions to determine whether the biological conditions found during the
study confonn to expectations based on the existing model relations.
When appropriate information from previous studies is not available for relevant
comparisons, new information on which to complete relative assessments of habitat to
biological condition relationships can be developed by conducting physical, chemical,
and biological studies in one or more reference waters using the same methods used in
the study waterbody. Including three, five, or more reference waters in these efforts
allow developing new quantitative relationships of habitat to biological conditions
relationships in the references waterbodies. These results can then be used to assess
habitat and biological conditions in the study waterbody of concern. Use of fewer
reference waterbodies allows less rigorous and sometimes only qualitative comparisons
to be completed between the reference and study waterbodies. This section briefly
introduces sources for information on the appropriate methods and models available to
evaluate aquatic habitat quality.
6.4.2.1 U.S. EPA Use Attainability Analyses
EPA's three Technical Support Manuals: Waterbody Surveys and Assessments for
Conducting Use Attainability Analyses provide useful guidance on methods to measure
and assess individual habitat variables and on approaches to integrate these measures
to assess habitats in streams and rivers (U.S. EPA, 1983b), estuaries (U.S. EPA, 1984a),
and lake and reservoirs (U.S. EPA, 1984b).
Habitat procedures discussed in the technical support manual for flowing waters
(U.S. EPA, 1983b) include the USDI Fish and Wildlife Services Habitat Suitability Index
Models (see Section 6.3.9;3), a broad selection of diversity indices and measures of
community structure, a community recovery index, intolerant species analyses,
omnivore-carnivore (trophic structure) analysis, use of reference sites, and important
considerations to aid interpretation of these measures with analyzing attainable uses.
For lake and reservoir, the technical support manual (U.S. EPA, 1984b) includes
discussions of routine physical habitat measurement techniques; nutrient models
(primarily phosphorus loading models); hydrodynamic circulation models; general
environmental relationships for phytoplankton, aquatic plants, zooplankton, benthic
invertebrates, and fish; chlorophyll and trophic state models; indicator organisms; use
6-53
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
of reference site information; aquatic life rating systems; and remediation and restoration
techniques.
Biological conditions in estuaries and related near-shore brackish water
environments is principally controlled naturally by salinity and water circulation patterns
that affect salinity, as discussed in Section 6.2.5.4. U.S. EPA's (1984a) technical support
manual for conducting use attainability analyses in estuaries discusses physical processes
affecting estuarine processes; estuaiine classification systems; models for assessing the
influences by physical and chemical conditions on use attainability; the plants and
animals specially adapted to allow inhabiting estuaries; habitat suitability models for
estuarine species; and approaches for synthesizing and interpreting monitoring results.
6.4.2.2 U.S. EPA Rapid Bioassessment Procedures
EPA's Rapid Bioassessment Protocols for Use in Streams and Rivers (Plafkin et al.,
1989) provides a systematic approach, including five individual protocols, for integrating
field data for habitat conditions with data for benthic macroinvertebrates and fish to
assess the quality of flowing water systems. These protocols were developed under the
approach that the overall assessment of ecological conditions must first focus on
evaluating habitat quality, then aquatic system conditions represented by biological data
can be analyzed in light of the defined habitat quality. In other words, habitat, as the
principal determinant of biological potential, sets Hie context for interpreting biological
survey or monitoring results; habitat can be used as a general predictor of biological
conditions. Routine water chemistry can also help to characterize certain impacts.
In the Rapid Bioassessment Protocols (RBPs) I and IV, habitat evaluations are
emphasized in the final assessments because of minimal efforts spent in these
approaches in collecting and analyzing biological data. In RBPs II, in, and V, however,
the biological evaluations are more rigorous and appropriately take prominence in the
evaluations. Habitat assessments support the evaluations in these three protocols. It can
identify obvious constraints on the attainable potential of the assessed waterbodies, help
in selecting appropriate sampling stations, and provide basic information for interpreting
the results from the biological samples. This document provides detailed guidance on
methods appropriate for sampling, measuring, and assessing habitat and biological
population in flowing waters; and evaluating the relationships between habitat quality
and biological conditions. In discussing how to complete integrated assessments, six
main topics are considered:
• Evaluating habitat at a site-specific control relative to that at a regional
reference.
6-54
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Evaluating water quality effects.
Evaluating biological impairment due to reversible habitat alterations.
Evaluating an alternative site-specific control station.
Bioassessment using a site-specific control station.
Bioassessment using a regional reference.
6.4.2.3 USDI FWS Habitat Suitability Index (HSD Models
The Habitat Suitability Index (HSI) Models developed by the U.S. Fish and
Wildlife Service provide another source of potentially useful information to help
interpret relationships between habitat conditions and resident biota in receiving waters.
These models are synthesized from facts, ideas, and concepts obtained from the research
literature and expert reviews.
For the most part, HSI models are uncalibrated, unverified, and {invalidated.
Consequently, developers and users generally acknowledge the limitations on the
predictive accuracy of these models. Nonetheless, HSI models provide valuable
compilations of the literature for each species on which HSI reports have been prepared.
Further, die models themselves target the habitat variables most generally found to have
high relationships with die modeled species and the relative weights given each habitat
variable included in each model provide a useful guide to die relative importance of
each variable to the fish species. In essence, these models provide well founded
hypotheses on specific species-habitat relationships. Table 6-8 lists the available HSI
models for various aquatic species and sources for copies of these models.
6.4^.4 Regression based stream habitat-fish models
Much research has focused on relationships among stream fisheries, habitats, and
flows. This research is reviewed by Loar and Sale (1981), Fausch et al (1988), and
others. Fausch et al. (1988), for example, reviewed 99 instream flow and habitat quality
models. Most of these models were developed using regressions methods to define the
relationship of measured habitat characteristics to fish numbers or densities in streams.
The authors and other similar reviewers conclude that there are many potential problems
in all of these models and die should be used with caution. In some applications of die
TMDL framework, there will arise opportunities to use of these models to help define
potential uses of the receiving water by fish. When these occasions arise, it is important
6-55
-------
The Cadmus Group, Inc. DRAFT June 8,1993
Table 6-8. Habitat Suitability Index Models for Freshwater Fish, Saltwater Fish,
Aquatic Invertebrates, and Aquatic Vertebrates
Taxonomlc Family
Common Name
Scientific Name
NTIS Number1
USDI PWS Number1
FRESHWATER FISHES
Acipcntcridae
Shortnoae Sturgeon
Athcrinidae
Inland Silverside
Polyodontida*
Paddlefiah
Saloionidac
Arctic Grayling
Brook Trout
Brown Trout
Chinook Salmon
Coho Salmon
Pink salmon
Chum salmon
Cutthroat Trout
Lake Trout
Rainbow Trout
Eaocidae
Northern Pike
Muskellung*
Cyprinidae
Blacknose Dace
Longnoae Dace
Common Carp
Common Shiner
Creek Chub
Pallfish
CatotfomMae
Bigmouth Buffalo
Smallmouth Buffalo
Longnose Sucker
White Sucker
Aapaaer bremrostrum
Menidia beryilina
Potydon rpathula
TkymaUut arctiau
Satodinus fontinaiit
Salmo trutta
Oncorkyndnu tthmyttcha
Oncorftynchus tisutch
Oncorkyndm gorbuscha
Otworftyruhu* krta
Onehehyndna dark)
Snivel inm namoycuth
Oncorhyndwt myUu
Esac luctus
Exa moKfuitwngy
FUtinichthyt atratuiui
FMnickthys cataract*
Cypiinut earpk)
Notropis comutut
SemotSus atrmaaibtus
SemotHus aorporalii
Ictioina cyprmeiha
Ictiobu* bubalus
Catostomus catostomus
Catostomut commertoni
PB83-147M1/AS
PB86-1260Q2/AS
PB82-239922/AS
PB84-129501/AS
PB84-150615/AS
PB84-128529/AS
PB82-239914/AS
FWS/OBS-82/10.129
FWS/OBS-82-10.80
FWS/OBS-82/10.80
FWS/OBS-82/10.110
FW5/OBS-82/10.24
FWS/OBS-82/10.124
FWS/OBS-82/10.122
FWS/OBS-82/10.49
FWS/OBS42/10.109
FWS/OBS-82/10.108
FWS/OBS-82/10.5
FWS/OBS-82/10.84
FWS/OBS-82/10.60
FWS/OBS-82/10.17
FWS/OBS-82/10.148
FWS/OBS-82/10.41
FWS/OBS-82/10.33
FW5/OBS-82/10.12
FWS/OBS-82/10.40
FWS/OBS-82/10.4
FWS/QBS-82/1048
FWS/OBS-82/1034
FWS/OBS-82/10.13
FWS/OBS-82/10.33
FW5/OBS-S2/10.64
6-56
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Taxonomic Family
Scientific Name
NT1S Number1
USDI FWS Number2
Common Name
Ictahiridae
Channel Catfish
Ictalurus punctatus
FWS/OBS-82/10.2
Flathead Catfish
Pylodkta o/iwrij
FWS/OBS-82/10.152
Blade Bullhead
Ictolunu mehu
PB83-147025/AS
FWS/OBS-82/10.H
Pereichthyidaa
Striped Baas (Inland)
Morone aaxatUis
FWS/OBS-82/10A5
Striped Baas (Coastal)
Monme aaxatttu
FWS/OBS-82/10.1
White Baaa
Morone chrysops
FWS/OBS-82/10.89
Centnurhidae
Black Crappie
Pamoacit nigromamltttut
PB82-239930/AS
FWS/OBS-82/10.6
BluegUl
Lrporna maovchtna
FWS/OBS-82/10.8
Green Sunfiah
Lepomis cyatuDus
FWS/OBS-82/10.15
Redbreast Sunfiah
Lepomis tzuritut
FWS/CMS-82/10.119
Redear Sunfiah
Lepomis microbptou
FWS/OBS-82/1079
Largemouth Ban
Micmpterus jalmoida
FWS/OBS-82/10.16
Smalhnouth Ban
Miavptena dolomieui
FWS/OBS-82/1036
Spotted Baaa
Micwpterua punctulatus
FWS/O8S-82/1072
Warmouth
Lepomh gukmt
FWS/OBS-82/10.67
White Crappie
Pomocri* mnuiari*
FWS/OBS42/107
Clnpeidae
Alewife Bhiebeck Herring
A1(WI
PB8S-222699/AS
FWS/OBS-82/1038
American Shad
Alan ttrpidiuton*
FWS/OBS-82/10.88
Gizzard Shad
LAJ9U6UJM CeptatBnUm
PB86-113586/AS
FW5/OBS-82/10.112
Perddae
Slough Darter
Ethontoma gncUe
FWS/OBS-82/10.9
Walleye
StizoUedkm vtinwtt
FWS/OBS-82/1046
Yellow Perch
Perm fkvaans
PW5/OBS42/1055
SALTWATER FISHES
Sciaenidae
Spotted Seatrout
Cynacion mbvkma
FWS/OBM2/10.75
Southern Kingfish
Mtntidrrkta amtricamu
PB85-222602/AS
FW5/OBS-82/10-31
juvenile Atlanta Croaker
Mkrvpogonk* wtduktus
FWS/OBS-82/10.98
(Revised)
Juvenile Spot
Lekmtomut xonthvrus
PB83-148197/AS
FWS/OBS-82/10.20
Red Drum
Scmnopt oaMettu
FWS/OBS-82/10J4
(Larval and Juvenile)
6-57
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Taxonomk Family
Scientific Name
NTIS Number1 USDIFWS Number1
Common Name
Bothidae
Southern and
Paralichthyi albigutta
FWS/OBS-82/10.92
Gulf Flounders
Paralichtkys lethoetigm
Clupeidae
Gulf Menhaden
Brevoortia patrmus
PB83-142513/AS FWS/OBS-82/10.23
Pleuronectidae
Juvenile English Sole
Parophrys vetuius
FWS/OBS-82/10.133
AQUATIC INVERTEBRATES
Pink Shrimp
Pauma duorarum
FWS/OBS-82/10.76
Hard Clam
Mercenaria mereenam
FWS/OBS-82/1077
Red King Crab
Paralithoda camttchatica
FWS/OBS-82/10.153
AQUATIC VERTEBRATES
Slider Turtle
Pseudanyt scripts
FWS/OBS-82/10.125
Snapping Turd*
Quhfdm wpentina
FWS/OBS-82/10.141
American Alligator
Alligatorulat mistissipprntu
FWS/OBS-82/10.136
Bullfrog
Ram cateibinma
FWS/OBS-82/10.138
1 USFWS out-of-print publications available from the National Technical Information Service (NTI5), US.
Department of Conurwroe, S285 Port Royal Road, Springfield, Virginia 22161.
1 Single copies available from the US. Pish and Wildlife Service, Publications Unit Arlington Square Building,
Mail Stop 725, 18th and C Streets NW, Washington, DC 20240.
to be aware of potential limitations and the necessary cautions when using these models.
This subsection reviews the development of these models, their potential shortcomings,
and necessary considerations for their subsequent use. This discussion draws heavily
from the reviews of Fausch et al. (1988) and Marcus et al. (1990).
In concept, most habitat investigations endeavor tcr develop techniques and
models through which standing crops and/or other measures of biological productivity,
generally as pertaining to fish, can be described or predicted using a set of habitat
variables. Underlying all of the resulting models is die premise that for each habitat
variable, or set of habitat variables, there are definable limits, beyond which conditions
become unsuitable for fish. Somewhere between these upper and lower extremes,
optimal conditions exist that grade predictably to the unacceptable conditions.
6-58
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Ultimately, after appropriate relationships are derived, it is hoped that a few well
chosen, easily obtained measurements made for a stream can be entered into a model
to predict the stream's potential carrying capacity for fish.
Present models, which have been developed using both qualitative and
quantitative approaches, include as few as one to as many as 21 input variables. Some
variables included in some models are transformed from or derived (recombined) using
the originally measured variables. Overall, the variables include details on basin
morphology, channel morphology, flow rates, habitat structure, species present, and
other physical and chemical measurements.
Among the assorted problems are associated with all existing habitat models, the
potentially most damaging in the long term is that no truly standard methods exist for
measuring habitat variables. Without use of consistent measurements, it is impossible
to compare or synthesize data from different investigators. Moreover, because methods
used in collecting data are reflected in the resulting models developed, single data sets
can not be explored using otherwise similar models that are based on alternative
sampling methods.
Various problems associated with many of the regression based models have
statistical bases. First, presentations for most of these models lack information necessary
to critically evaluate how the model was statistically selected or how the model may
perform in general application. To evaluate the statistical worth of models, sufficient
information should be included to enable evaluation (1) die correlation coefficients (r);
(2) coefficients of determination (i.e., r2, R2, or adjusted R2: R2 is the r2 for multiple
regression relationships and adjusted R2 is corrected for the number of variables
included in die equation); (3) standard errors for the regression coefficients (i.e., is the
coefficient for each variable included in the model significantly different from 0); and/or
(4) the confidence interval for the presented models.
Another statistical problem in many of the models is small sample sizes. This
potentially limits the applicability of die model to the limited ranges in habitat variability
use to develop the model If models are used to extrapolate outside these limits, the
resulting predictions can be biased and unreliable. Subsequent model calibration and
verification, as discussed in Section 7.1, can help to overcoming this limitation.
Further, errors associated with measuring the various habitat variables have rarely
been evaluated during model development. If measurements upon which the model is
based are biased, the model will yield similarly biased predictions. Most models have
not been tested with data that was not used in developing the models. Thus, we
generally know little of the overall realism, precision, or generality of the models.
6-59
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Various potentially unreasonable assumptions about habitat relationships are also
implicit in many of these models. For example, the U.S. Fish and Wildlife Service's
Instream How Incremental Methodology (IFJM) include potentially erroneous
assumptions that (1) fish primarily respond to average water velocities at some defined
depth (e.g., 0.6 of the depth below the surface); (2) stream depths, velocities, and
substrates are not related to each other (i.e., an underlying assumption for regression
analysis is that independent variables are uncorrected); and (3) large amounts of
suboptimal habitat are equivalent to a small amount of optimal habitat Yet, studies
show that fish respond more to flow difference in microhabitats; that stream depths,
velocities, and substrates are often highly correlated; and that suboptimal habitats can
often be uninhabitable.
Most models include at least one additional and perhaps also unreasonable
assumption about the relationship between measured habitat variables and measured
fisheries densities or biomasses used to derive the models. That is, it is often assumed
that the measured density or biomass for fish is at the carrying capacity for the habitat,
and that this carrying capacity is defined by those physical and chemical conditions
measured in the habitat This assumption precludes such effects as predation (including
fishing), or competition as having any potential influence on the population.
Finally, while a plethora of models relating stream habitat to fish are indeed
available, few models, it any, currently available has been verified as reliably predicting
effects of stream flow alterations on standing crops of fish.
6.5 Data Management Techniques
6.5.1 Data management
All monitoring data should be organized and stored in a form that allows ready
access. The voluminous and diverse nature of the data, and the variety of individuals
who can be involved in collecting, recording and entering data, can easily lead to the
loss of data or the recording of erroneous data. Lost or erroneous data can severely
damage the quality of monitoring programs. A sound and efficient data management
program for a monitoring program should focus on preventing such problems. This
requires that data be managed directly and separately from the activities that use them.
Data management systems comprise technical and managerial components. The
technical components involve the selection of appropriate computer equipment and
software and the design of the database, including data definition, data standardization,
and a data dictionary. The managerial components include data entry, data validation
and verification, data access, and methods for users to access the data.
6-60
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
To ensure the integrity of the database, it is imperative that data quality be
controlled from the point of collection to the time the information is entered into the
database. Field and laboratory personnel must carefully enter data into proper spaces
on data sheets and avoid transposing numbers. To avoid transcription errors, entries
into a database should be made from original data sheets or photocopies. As a
preliminary screen for data quality, the database design should include automatic range-
checking of all parameters. Values outside the defined ranges should be flagged by the
program and immediately corrected or icluded in a follow-up review of the entered
data. For some parameters, it might be appropriate to include automatic checks to
disallow duplicate values. Preliminary database files should be printed and verified
against the original data to identify errors.
Additional data validation can include expert review of the verified data to
identify possible suspicious values. Sometimes, consultation with the individuals
responsible for collecting or entering original data is required to resolve problems. After
all data are verified and validated, they can be merged into the monitoring program's
master database. To prevent loss of data from computer failure, at least one set of
duplicate (backup) database files should be maintained at a location other than where
the master database is kept
6.5.2 Record Keeping Requirements
NPDES permits typically require that all data collected to comply with permit
conditions be retained for at least five years from the date of sampling, recording, or
permit application. Information that must be retained includes calibration and
maintenance records,, strip chart recordings from continuous monitoring equipment,
reports, and data records. Data records should include the following information;
• The date, location, and time of sampling or measurements;
• The persons who performed the sampling;
• The date analyses were performed
• The persons who performed the analyses;
• The analytic techniques used; and
• The results of the analyses.
6-61
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
6.5.3 GIS and Data Analysis in TMDL Planning
6.5.3.1 Introduction to GIS
TMDL development frequently requires analysis of data over broad spatial scales
(e.g., across a watershed), utilizing land attributes such as basin size, land use types,
slopes, and vegetation cover. A geographic information system (GIS) can be an integral
part of this process, depending on the phase of TMDL development, the type of loading
model vised, and the resources available. GIS is being used in a growing number of
fields, and it is likely that those developing TMDLs will have access to a GIS currently
or in the near future. The application of GIS in the TMDL process can include data
management, the presentation of data, and the interpretation or analysis of data. The
most significant use of GIS is as a sophisticated and powerful decision support tool to
analyze and manipulate data. GIS allows the examination of relationships among data
that would otherwise be too complex or cumbersome to discern. The TMDL process is,
by necessity, one that must take a "big picture" view of water quality problems. GIS is
well suited to this scale of analysis, and there are many possible uses for GIS in this
process.
In addition to representing data in the form of maps, many GIS provide a set of
tools for the analysis of data. Analysis may be in the form of overlay operations (i.e.,
analogous to using mylar map overlays), aggregating data to a desired level, or statistical
functions. It is also possible to implement algorithms and automate procedures in many
GIS using a "macro" programming language. The type of GIS analysis being utilized
will correspond to the level of detail inherent in die different phases of TMDL
development In following with the rest of this guidance document, this section will
address the use of GIS for: 1) gaining an understanding of water quality issues in an
area, 2) performing a scoping or screening analysis to prioritize issues, and 3) detailed
modelling relevant to TMDL development While the first two purposes of GIS will be
discussed briefly, the focus of this section will be on the use of GIS in relation to
modelling for the TMDL process.
6.5.3.2 Use of GIS
At a general level, a GIS is a system for the collection, storage, manipulation, and
display of geographically referenced data. It is a set of tools that has evolved from
various fields (e.g., cartography, geography, and computer science), and there are several
different GIS currently available. There are two basic types of GIS: a raster-based system
or a vector-based system. A raster-based system represents data in a grid format, in
which each cell of the grid has a single value for the attribute in question. A vector-
based GIS maintains the topological relationships between geographical objects, which
6-62
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
are stored as points, lines, or polygons. A discussion of the technical details of these two
types of GIS is out of the scope of this document, and the reader should refer to a GIS
text for greater detail. Historically, vector-based GIS have tended to be used in the
planning field, and raster-based GIS have been used for scientific analyses. Over time,
this distinction has become much less clear, and there are now GIS that can handle data
in both raster and vector formats.
GIS software can run on a variety of hardware platforms, ranging from personal
computers to powerful workstations. The more sophisticated GIS packages offer a great
deal of functionality, but this is often offset by die level of complexity in using the GIS.
Providers of GIS are currently addressing this issue, and more menu-driven, user-
friendly GIS are becoming available. Regardless of the GIS in question, it can aid in the
first two phases of the TMDL process (Le., gaining an understanding of water quality
issues in an area and performing a scoping or screening analysis to prioritize issues).
Using GIS in combination with water quality models, or implementing the model
through GIS itself, will depend much more on the specifics of the GIS package and the
model chosen.
The TMDL process has evolved from examining individual stretches of rivers or
portions of basins to include entire watersheds. Because there can be several water
quality problems that require attention in a particular watershed, gaining an
understanding of the water quality issues in a watershed is frequently die starting point.
Given the appropriate data, or even data of limited detail (e.g., approximations of slope
and crude delineation of land uses), GIS can quickly supply information to decision-
makers about the nature of the problems at hand. If there is a particular water quality
issue being addressed, GIS can he used to gain an understanding of the contributing
factors and to indicate the next step in addressing the issue. Used in this way, GIS can
provide a context for further decisions and analyses in the TMDL development process.
Performing scoping and screening analyses with a GIS can improve and speed up
this step in TMDL development, although the data requirements tend to be greater. In
general, data gathering and data input are the most time- and effort-intensive aspect of
a project using GIS technology; however, once the necessary data are in a GIS, it is
relatively simple to use these data in a meaningful way. For example, by assigning
nutrient or sediment export coefficients to different land uses, GIS could be used to
screen out priority areas (e.g., those near water bodies) or to estimate the amount of
pollutant reduction that is necessary (e.g., across agricultural regions). Based on the
spatial characteristics of an area, GIS can help to break a basin up into hydrologic sub-
units or to isolate those with specific attributes. Many other analyses, such as buffering
riparian areas and accounting for distances between stream gauges, are also possible
through GIS. These types of analyses do not require extensive GIS training to
6-63
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
accomplish, and their applicability is obvious. Without GIS, many analyses would
require too much time or would not be possible without a great deal of experience.
After prioritization of water quality issues, GIS can facilitate detailed modelling
relevant to the TMDL process. Loading can be estimated by models that range in
complexity from an equation implemented on a hand-held calculator to sophisticated,
computer-based systems. As was discussed earlier, a GIS is usually a raster- or vector-
based system. The choice of a water quality model, and its corresponding requirements,
may largely determine whether a raster- or vector-based GIS is preferred. On the other
hand, much of the data required to develop a TMDL may already be in an existing GIS.
If this is the case, the availability and format of die data itself may help in choosing
among the many water quality models in existence.
6.5.3.3 Linkages between GIS and TMDL Models
There are many ways of incorporating GIS into the modelling of water quality.
Some models have "seamless" linkages (i.e., transparent to the user) with certain GIS
packages, but most do not. Frequency, a GIS can provide summary statistics and other
input parameters, depending on die model used and the data available. Therefore, the
link between GIS and TMDL models can be automated to different degrees, or not at all.
When compared to providing parameter data manually (e.g., aggregating classes of data
or calculating the coincidence of several different attributes), GIS is of great value for this
task. GIS is also well-suited to displaying the output of models, provided that the
output is geographically referenced in some way, and visualizing various mitigation
scenarios can aid in their evaluation.
Table 6-9 summarizes many of the data inputs for watershed-scale models. It is
clear that commonly available GIS data, which are stored in "layers" that represent
different themes or land attributes (e.g., land use types, soils, vegetation cover, and
slope), can be manipulated to meet many of the data needs listed in this table. Again,
note that the linkage between a GIS and a particular model can be automated to various
degrees, and it may also go the other way (i.e., output from a model into a GIS to
display modelling results). Extracting input parameters from a GIS may involve several
GIS operations, and the subsequent input of these parameters into a model may require
that they be entered manually. In contrast, resources and programming expertise may
be available so that this is accomplished through a few commands. Regardless of how
sophisticated this link may be, GIS has been able to improve and facilitate the water
quality modelling process.
6-64
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Table 6-9. Input Data Needs for Watershed Models
1. System Parameters:
Watershed size
Subdivision of the watershed into homogenous subareas
Imperviousness of each subarea
Slopes
Fraction of Impervious areas directly connected to a channel
Maximum surface storage (depression plus interception storage)
Soil characteristics including texture, permeability, erodibility, and composition
Crop and vegetative cover
Curb density or street gutter length
Sewer system or natural drainage characteristics
2. State Variables
Ambient temperature
Reaction rate coefficients
Adsorption/desorption coefficients
Growth stage of crops
Daily accumulation of rates of litter
Traffic density and speed
Potency factors for pollutants (pollutant strength on sediment)
Solar radiation (for some models)
3. Input Variables
Precipitation
Atmospheric fallout
Evaporation rates
Source: U5. EPA, 1992a (after Novotny and Chester*, 1981)
The U.S. Army Corps of Engineers Hydrologic Engineering Center (HEC) has
pursued the integration of hydrologic modeling and GIS extensively. A recent review
of GIS applications in hydrologic modeling is provided by DeVantier and Feldman
(1993), ASCE Journal of Water Resources Planning and Management (Vol. 119, No. 2,
March/April 1993 contains a special issue focusing on GIS applications and covering
topics such as nonpoint pollution and urban stormwater management.
6-65
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
6.5.3.4 Other Uses of GIS in the TMDL Process
As was mentioned above, some water quality modelling has been implemented
through GIS alone, without the use of a separate model such as those described above.
This use of GIS may require significant programming and GIS expertise, but the
resulting models can be specific to the geographic characteristics of the area and the
needs of those performing the analysis. Linking GIS to other types of models (e.g.,
groundwater flow models) can also yield insights to water quality that would not
otherwise be possible. These approaches may be most appropriate in areas where water
quality analyses cannot be performed using existing water quality models. The
following two case studies illustrate these uses of GIS.
"Land Use Change and Impacts on the San Francisco Estuary; A Regional
Assessment with National Policy Implications"
Water quality modelling over very large areas, such as that contributing to major
estuaries, may require model implementation through GIS alone. McCreary et al.
(1992) describe the use of GIS in modelling changes in water quality and losses
of different wetland types, given development scenarios in the areas affecting the
San Francisco estuary. Runoff was estimated from local precipitation averages
and the imperviousness of different land use types. Contaminant concentrations,
which included metals, nutrients, biological oxygen demand, and total suspended
solids, were derived from national and local studies.
"Analyzing Septic Nitrogen Loading to Receiving Waters Waquoit Bay,
Massachusetts"
This case study by Sham et al. (1993) shows how GIS can be linked to a
groundwater model to analyze changes in water quality over time. By
incorporating the spatial and temporal characteristics of development in the
watershed, title accuracy of nutrient loading estimates can be improved. This
study indicates how important a role groundwater can play in water quality
problems and how a long-term view must be taken in their solution.
One way to ensure the success of a GIS operation is to integrate it into as many
tasks as possible. In addition to water quality modelling, GIS can also play a part in the
management of data related to water quality and the TMDL process. Because these data
have a spatial component, GIS can be an excellent facility for their management, analysis,
and display, as the next two studies indicate.
"Application of a GIS to a Water Regulatory Permit Inventory"
This case study (Crowell, 1989) describes how the Southwest Florida Water
Management District has used GIS to inventory and update permit information,
instead of performing the process on USGS topographic maps. Regulatory permit
6-66
-------
The Cadmus Group, lite.
DRAFT
June 8,1993
mapping (and the storage of relevant permit data) via GIS has not only aided in
the evaluation of permit applications, but it has streamlined data management
and limited access to only the most up-to-date mapped permit information.
"Geographic Information Systems as a Tool in Water Use Data Management"
Schoolmaster and Marr (1992) discuss some of the possible uses of GIS in the
management of water use data, incorporating spatial and temporal dimensions
for both surface and groundwater. While this study does not specifically address
water quality issues, it does discuss concepts for data storage, use, and display
that could be useful for managing data in the TMDL process.
6.6 Quality Assurance and Quality Control—(QA/QC)
QA/QC procedures are essential to ensure that data collected in environmental
monitoring programs are useful and reliable. Quality assurance refers to programmatic
efforts to ensure the quality of monitoring and measurement data. Quality control, which
is a subset of quality assurance, refers to the routine application of procedures designed
to obtain prescribed standards of performance in monitoring and measurement. This
section introduces procedures for sample and analytic quality control and for field
quality assurance. U.S. EPA (1985) and Plafkin et al (1989) provide additional
information for defining QA/QC program plans as part of the monitoring design
process, including the development of QA/QC adequate program descriptions,
integration QA/QC programs with monitoring program plans, EPA's responsibility in
the QA/QC process, and the importance of QA/QC in the bioassessment process.
6.6.1 Sample and Analytic Quality Control
The following techniques are useful in assessing sampling and analytic
performance (see also U.S. EPA, 1979b):
• Duplicate samples. Duplicate samples collected at selected locations using
two sets of field equipment or by grab sampling provide a check for
precision in sampling equipment and techniques.
• Split samples. Single samples split and analyzed separately check for
variation in laboratory method or between laboratories. Samples can be
split and submitted to a single laboratory or split between laboratories.
6-67
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Spiked samples. Introduced a known quantity of a substance into a volume
of distilled water and analyzing for that substance provides a check of the
accuracy of laboratory and analytic procedures.
Reagent blanks. Preserving and analyzing a quantity of distilled water in
the same manner as environmental water samples can indicate
contamination caused by sampling and laboratory procedures. Normally
the value of pollutant measured in the blank is subtracted from the values
obtained by analyzing environmental samples.
6.6.2 Field Quality Assurance
Errors or a lack of standardization in field procedures can significantly decrease
the reliability of environmental monitoring data. A quality assurance plan for field
measurement procedures and equipment should at a minimum include the following
elements:
• Identification of the analytic method, including special handling
procedures. In most cases analytic methods will be specified by regulation.
• Allocation of field and laboratory analyses to quality control.
• Procedures for recording, processing, and reporting data.
• Procedures for and records of maintenance and calibration of field
instruments.
• Evaluation of the performance of field personnel.
It is important that quality procedures be followed and regularly examined. For
example, field meters can provide erroneous values if they are not regularly calibrated
and maintained. Reagent solutions and probe electrolyte solutions have expiation
periods and should be refreshed periodically.
6-68
-------
The Cadmus Group, Inc.
DRAFT
June 8; 1993
References. Chapter 6
Adams, W.J. 1992. Sediment Quality and Aquatic Life Assessment. Environmental
Science and Technology,26(10):1864~1875.
American Public Health Association (APHA). 1992. Standard Methods for the Examina-
tions of Water and Waste Water, 18th edition. American Public Health
Association, Washington, DC.
American Society for Testing and Materials (ASTM). 1991. Standard Guide for
Conducting Sediment Toxicity Tests with Freshwater Invertebrates. ASTM E-
1383-94. American Society for Testing and Materials, Philadelphia, PA.
Burton, G.A. 1992. Sediment Toxicity Assessment. Lewis Publishers, Boca Raton, FL.
Crowell, M.L. 1989. Application of a GIS to a water regulatory permit inventory, in
GIS/US '89 Proceedings, Orlando, FL, November 26-30,1989, Volume 2, pp. 669-
678.
Dennis, W.M. and B.G. Isom (editors). 1984. Ecological Assessment of Macrophyton:
Collection, Use, and Meaning of Data. ASTM Special Technical Publication 843.
American Society for Testing and Materials, Philadelphia, PA.
Devantier, B.A. and A.D. Feldman. 1993. Review of GIS applications in hydrologic
modeling. Journal of Water Resources Planning and Management, 119 (2):246-261
Dunne, T., and L.B. Leopold. 1978. Water in Environmental Planning. W.H. Freeman
and Company, San Francisco, CA.
Everhart, W.H., A.W. Eipper and W.D. Youngs. 1975. Principles of Fishery Science.
Cornell University Press, Ithaca, NY.
Fausch, K.D., C.L. Hawks, and M.G. Parsons. 1988. Models that Predict Standing Crop
of Stream Fish from Habitat Variables (1950-1985). PNW-GTR-213. U.S.
Department of Agriculture, Forest Service, Pacific Northwest Research Station,
Portland/ OR.
FHWA, 1990. Pollutant Loadings and Impacts from Highway Stormwater Runoff,
Volume I: Design procedure. FHWA-RD-88-007.
Haith, D. A. and L. L. Shoemaker. 1987. Generalized watershed loading functions for
stream flow nutrients. Water Resources Bulletin 23(3):471-478.
6-69
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
ISCO. 1985. Open Channel Flow Measurement Handbook. Lincoln Nebraska.
Klemm, D.J., P.A. Lewis, F. Fulk, and J.M. Lazorchak. 1990. Macroinvertebrate Field
and Laboratory Methods for Evaluating the Biological Integrity of Surface Waters.
EPA/600/4-90-030. Environmental Monitoring Systems Laboratory, U.S.
Environmental Protection Agency, Cincinnati, OH.
Lagler, K.F. 1956. Freshwater Fishery Biology. Wm. C. Brown Company, LA.
Lind, O.T. 1985. Handbook of Common Methods in Limnology. Kendall/Hunt
Publishing Company, Dubuque, LA.
Loar, J.M., and M.J. Sale. 1981. Analysis of Environmental Issues Related to Small Scale
Hydroelectric Developments. V: Instream Flow Needs for Fisheries Resources.
ORNL/TM-7861, Oak Ridge National Laboratory, Oak Ridge, TN.
Lowe, R.L. 1974. Environmental Requirements and Pollution Tolerance of Freshwater
Diatoms. EPA-670/4-74-005. National Environmental Research Center, U.S.
Environmental Protection Agency, Cincinnati, OH.
Marcus, M.D., M.K. Young, L.E. Noel and B.A. Mullan. 1990. Salmonid-habitat
Relationships in the Western United States: a Review and Bibliography. GTR
RM-88. USDA Forest Service, Rocky Mountain Forest and Range Experiment
Station, Fort Collins, CO.
McCreary, S., R. Twiss, B. Warren, C. White, S. Huse, K. Gardels, and D. Roques. 1992.
Land use change and impacts on the San Francisco estuary: A regional assessment
with national policy implications. Coastal Management, 20, pp. 219-253.
Merritt, R.W. and K.W. Cummins. 1984. An Introduction to the Aquatic Insects of
North America, 2nd ed. Kendall/Hunt Publishing Company, Dubuque, LA.
Mount, D.I. 1989. Methods for aquatic toxicity identification evaluations. Phase m
toxicity confirmation procedures. EPA/600/3-88/036. U.S. Environmental
Protection Agency, Duluth, MN.
Mount, D.I. and L. Anderson-Carnahan. 1988. Methods for aquatic toxicity
identification evaluations. Phase I toxicity characterization procedures.
EPA/600/3-88/034. U.S. Environmental Protection Agency, Duluth, MN.
6-70
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
Mount, D.I., and L. Anderson-Carnahan. 1989. Methods for aquatic toxicity
identification evaluations. Phase n toxicity identification procedures.
EPA/600/3-88/035. U.S. Environmental Protection Agency, Duluth, MN.
Nielsen, L.A., and D.L. Johnson (editors). 1983. Fisheries Techniques. American
Fisheries Society, Bethesda, MD.
Novotny, V. and G. Chesters, 1981. Handbook of Nonpoint Pollution: Sources and
Management Van Nostrand Reinhold, New York
Parmley, R.O. 1992. Hydraulics Field Manual. McGraw-Hill, New York.
Peltier, W. H. and C. I. Weber. 1985. Methods for measuring the acute toxicity of
effluents to freshwater and marine organisms. U.S. EPA, Environmental
Monitoring and Support Laboratory, Cincinnati, OH. Report No. EPA-600/4-85-
013.
Pennak, R.W. 1989. Fresh-water invertebrates of the United States, 3rd Edition. John
Wiley and Sons, Inc. New York, NY.
Plafkin, J.L., M.T. Barbour, K.D. Porter, S.K. Gross, and R.M. Hughes. 1989. Rapid
Bioassessment Protocols for Use in Streams and Rivers. Benthic
Macroinvertebrates and Fish. EPA/440/4-89/001. Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
Rantz, S. E. 1982. Measurement and computation of streamflow. Vol 1: Stage and
discharge; Vol. 2.: Computation of discharge. Water supply paper No. 2175. U.
S. Geological Survey, Washington, DC.
Ricker, W.E. 1975. Computation and interpretation of biological statistics of fish
populations. Bull. Fish. Res. Board Can. 191.
Schoolmaster, F.A. and P.G. Marr. 1992. Geographic information systems as a tool in
water use data management Water Resources Bulletin 28 (2): 331-336.
Schreck, C.B. and P.B. Moyle (editors). 1990. Methods for Fish Biology. American
Fisheries Society, Bethesda, MD.
Sham, C.H., J.W. Brawley, and MA. Moritz, 1993, "Analyzing Septic Nitrogen Loading
to Receiving Waters: Waquoit Bay, Massachusetts," upcoming in Proceedings of
the Second International Conference/Workshop on Integrating Geographic
6-71
-------
The Cadmus Group> Inc.
DRAFT
June 8,1993
Information Systems and Environmental Modeling, Breckenridge, CO, September
26-30,1993.
Thomann, R V., and J. A. Meuller. 1987. Principles of Surface Water Quality Modeling
and Control, Harper and Row Publishers, New York.
Thomas, N.A., J.O. Lamberson, and R C. Swartz. 1989. Chapter 2. In: Sediment
Classification Methods Compendium. Draft Final Report. Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1979a. Methods for the Chemical
Analysis of Water and Wastes. EPA-600/4-79-020. Environmental Monitoring
and Support Laboratory, U.S. Environmental Protection Agency, Cincinnati, OH
U.S. Environmental Protection Agency (U.S. EPA). 1979b. Handbook for Analytic
Quality Control in Water and Wastewater Laboratories. EPA-600/4-79-019.
Environmental Monitoring and Support Laboratory, U.S. Environmental
Protection Agency, Cincinnati, OH
U.S. Environmental Protection Agency (U.S. EPA). 1981. NPDES Compliance Flow
Measurement Manual. EPA MCD-77. Office of Water Enforcement and Permits
Enforcement Division, U.S. Environmental Protection Agency, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1983. Technical Support Manual:
Waterbody Surveys and Assessments for Conducting Use Attainability Analyses.
Office of Water, U.S. Environmental Protection Agency, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1984a. Technical Support Manual:
Waterbody Surveys and Assessments for Conducting Use Attainability Analyses,
Volume H: Estuarine Systems. Office of Water, U.S. Environmental Protection
Agency, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1984b. Technical Support Manual:
Waterbody Surveys and Assessments for Conducting Use Attainability Analyses,
Volume III: Lake Systems. Office of Water, U.S. Environmental Protection
Agency, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1988. NPDES Compliance Inspection
Manual Contract Nos. 68-01-6514,68-01-7050. Office of Water Enforcement and
Permits Enforcement Division, U.S. Environmental Protection Agency,
Washington, DC.
6-72
-------
The Cadmus Group, Inc.
DRAFT
June 8,1993
U.S. Environmental Protection Agency (U.S. EPA). 1989. Sediment Classification
Methods Compendium, Draft Final Report Office of Water, U.S. Environmental
Protection Agency, Washington, IX!.
U.S. Environmental Protection Agency (U.S. EPA). 1990. Draft Ecological Evaluation
of Proposed Discharge of Dredged Material into Ocean Waters. EPA/503/8-90-
002. Office of Marine and Estuaiine Protection, U.S. Environmental Protection
Agency, Washington, DC.
U.S. EPA. 1991a. Guidance for Water Quality-Based Decisions: The TMDL Process. EPA
440/4-91-001. OW/OST/OWOW
U.S. EPA. 1991b. PC Waterbody System User's Guide (Version 3.0). U.S. Environmental
Protection Agency, Office of Water, Assessment and Watershed Protection
Division, Washington, DC.
U.S. EPA. 1991c. Technical Support Document for Water Quality-Based Toxics Control.
EPA/505/2-90-001. Office of Water, U.S. Environmental Protection Agency,
Washington, DC.
U.S. EPA. 1991d. Assessment and Control of Bioconcentratabie Contaminants in
Surface Waters. Draft report. Office of Research and Development, Office of
Water, U.S. Environmental Protection Agency/ Washington, DC.
U.S. EPA. 1992a. NPDES Storm Water Sampling Guidance Document EPA 833-B-92-
001. Office of Water, U.S. Environmental Protection Agency, Washington, DC.
U.S. EPA. 1992b. Evaluation of Wet Weather Design Standards for Controlling Pollution
from Combined Sewer Overflows. EPA-230-R-92-006. Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
U.S. EPA, 1992c. Compendium of Watershed-scale Models for TMDL Development.
Report No. 841-R-92-002.
U.S. EPA. 1992d. Inventory of Exposure-Related Data Systems Sponsored by Federal
Agencies. EPA/600/R-92/07.
U.S. EPA. 1992e. Office of Water Environmental and Program Information Systems
Compendium. Office of Water EPA/800/B-92/001.
VanLandingham, S.L. 1982. Guide to the Identification, Environmental Requirements
and Pollution Tolerance of Freshwater Blue-Green Algae (Cyanophyta). EPA-
6-73
-------
The Cadmus Group; Inc.
DRAFT
June 8,1993
600/3-82-073. Environmental Monitoring and Support Laboratory, U.S.
Environmental Protection Agency, Cincinnati, OH.
Vollenweider, R.A. 1969. A manual on methods for measuring primary production in
aquatic environments. Blackwell Scientific Publications, Oxford and Edinburgh,
England.
Water Pollution Control Federation, 1983, Existing Sewer Evaluation and Rehabilitation,
Second Edition, Alexandria VA., Manual of Practice No. FD-6.
Weber, C.I. (editor). 1973. Biological Field and Laboratory Methods for Measuring the
Quality of Surface Waters and Effluents. EPA-670/4-73-001. National
Environmental Research Center, U.S. Environmental Protection Agency,
Cincinnati, OH.
Weber, C.I. et al. (editors). 1988. Short-Term Methods for Estimating the Chronic
Toxicity of Effluents and Receiving Waters to Marine and Estuarine Organisms.
EPA-600/4-87/02. Office of Research and Development, U.S. Environmental
Protection Agency, Cincinnati, OH.
Weber, C.I. et al (editors). 1989. Short-Term Methods for Estimating the Chronic
Toxicity of Effluents and Receiving Waters to Freshwater Organisms. EPA-600/4-
89/001. Office of Research and Development, U.S. Environmental Protection
Agency, Cincinnati, OH.
Weber, C.I. (editor). 1973. Biological Field and Laboratory Methods for Measuring the
Quality of Surface Waters and Effluents. EPA-670/4-73-001. National
Environmental Research Center, U.S. Environmental Protection Agency,
Cincinnati, OH.
Wedepohl, R.E., D.R. Knauer, G.B. Wolbert, H. Olem, P.J. Garrison, and K. Kepford.
1990. Monitoring Lake and Reservoir Restoration. EPA 440/4-90-007. Prepared
by the North American Lake Management Society for Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
Wetzel, R.G. and G.E. Likens. 1979. Limnological analyses. W.B. Saunders Company,
New York.
6-74
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
Chapter VII. Model Implementation, and Analysis
Purpose: Previous chapters have discussed the identification and selection of
simulation models for the estimation of TMDLs, and the development of monitoring
plans and data collection in support of modeling. This chapter discusses various issues
relating to the implementation of models, including the integration of modeling and
monitoring. Chapter 8 then gets to the basics of using modeling results to actually
perform TMDLs. Parts of the discussions contained in this chapter were adapted from
guidance for modeling and monitoring of combined sewer overflows (CSOs) presently
under development for OWEC.
7,1. Model Calibration and Validation
Once a model has been selected, model parameters are generally adjusted or fine-
tuned to reproduce observations at a particular site. The performance of the adjusted
model should then be checked for ability to predict on a separate set of data. These two
processes are known as model calibration and validation.
While the process in which the calibrated model is tested on an independent set
of data is commonly referred to in the modeling literature as "verification", it should be
noted that this usage is not really correct from a systems science perspective (see, for
instance, Nix et al., 1991). "Verification" has the implication of proving something to be
true. For environmental models we do not expect to be able to show that calibrated
model parameter values are exactly true and correct; instead, the best we can expect is
to show that the calibrated model does an adequate job of prediction. This is more
properly a test of the validity (rather than verity) of the calibration. Nix et al. suggest
that the term verification should be reserved for another step in the process: ensuring
that the computer code performs as expected. Therefore/ and in accordance with recent
guidance from the Office of Water (Martin et al., 1990), we will refer to the process of
testing the calibrated model as validation.
7.1.1 Role of Calibration and Validation
The purpose of model calibration and validation is to take a generalized numerical
construct or model, and, by specifying the 'right' parameter values (and using the 'righf
model subroutines) turn it into a site-specific predictive tool Calibration and validation
are thus integral parts of the model development process: When an established computer
model is used, calibration and validation can be thought of as the finishing touches
required to create the site-specific model (see Hgure 7*1).
7-1
-------
1
2
P
S3
£
o
1
2
<1
i
OS
£
* p*
CO
Figure 7-1. Principal Components of Modeling Framework
(after Thomann, 1980)
-------
The Cadmus Group> Inc.
DRAFT
June 10,1993
A TMDL developer could run a loading or receiving water model without
calibration for screening purposes, essentially to provide an informed "best guess" of
impacts, based in part on generic, rather than site-specific conditions. However, the
uncalibrated result is still only a guess, without confirmation. To use model simulation
results in support of a particular management alternative it is necessary to provide
evidence that the guess is a reasonable one. This is accomplished through the process
of model calibration and validation, which consists of fine-tuning the model to the site,
and proving that the results are reasonable (and measuring just how reasonable they
are). The process can be thought of as establishing the model's credibility as a witness.
The needs for, and functions of model calibration and validation are succinctly
summarized in Martin et al. (1990), which also provides extensive guidance on the
calibration/validation procedure as applied to WLA modeling of estuaries:
While models can be run with minimal data, their predictions are subject
to large uncertainty. Models are best operated to extrapolate from existing to
future conditions, such as in the projection of conditions under anticipated waste
loads. The confidence that can be placed on those projections is dependent upon
the integrity of the model, and how well the model is calibrated to that particular
[waterbody], and how well the model compares when evaluated against an
independent data set (to that used for calibration).
Model calibration is necessary because of the semi-empirical nature of
present day...water quality models. Although...formulated from the mass balance
and, in many cases, from conservation of momentum principles, most of the
kinetic descriptions in the models that describe the change in water quality are
empirically derived. These empirical derivations contain a number of coefficients
and parameters that are usually determined by calibration using data collected in
the [waterbody] of interest
Calibration alone is not adequate to determine the predictive capability of
a model... To map out the range of conditions over which the model can be used
to determine cause and effect relationships, one or more additional independent
sets of data are required to determine whether the model is predictively valid.
This testing exerdse...defines the limits of usefulness of the calibrated model.
Without validation testing, the calibrated model remains a description of the
conditions defined by the calibration data set The uncertainty of any projection
or extrapolation of a calibrated model would be unknown unless this is estimated
during the validation procedure.
Model calibration and validation interact strongly with both the data-collection
and model-selection activities of TMDL development covered in of this guidance. That
7-3
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
is, model calibration/validation requires data, and, conversely, the design of effective
monitoring plans can be guided by preliminary modeling results. Model development
and data collection thus interact in an iterative fashion (Figure 7-2). This will be true for
any effort involving application of other than the most simple models. However, it is
particularly appropriate to the development of TMDLs using the phased approach,
where ample time will be available for the collection of additional data and refinement
of modeling approach.
EPA's Compendium of Watershed-Scale Models for TMDL Development (Shoemaker
et al., 1992) summarizes the calibration and validation process as follows (we have
substituted "validation" for "verification" for consistency in usage with the current
document):
Calibration involves minimization of deviation between measured
field conditions and model output by adjusting parameters of the model
(Jewell et al., 1978). Data required for this step are a set of known input
values along with corresponding field observation results. The results of
[a] sensitivity analysis provide information as to which parameters have
the greatest effect on output. For the best results, CSO models should be
calibrated during storm events as opposed to dry flow periods (WPCF,
1989).
[Validation] involves the use of a second set of independent
information to check the model calibration. The data used for (validation]
should consist of field measurements of the same type as the data output
from the model. Specific features such as mean values, variability, extreme
values, or all predicted values may be of interest to the modeler and
require testing (Reckhow and Chapra, 1983). Models are tested based on
the levels of their predictions, whether descriptive or predictive. More
accuracy is required of a model designed for absolute versus relative
predictions. If the model is calibrated properly, the model predictions will
be acceptably close to the field observations.
[M]ost models are more accurate when applied in a relative rather
than an absolute manner. Model output data concerning the relative
contribution...to overall pollutant loads is more reliable titan an absolute
prediction of the impacts of one control alternative viewed alone. When
examining model output... it is important to note three factors that may
influence the model output and produce unreasonable data. First, suspect
data may result from calibration or [validation] data that are insufficient
or inappropriately applied. Second, any given model, including detailed
models, mav not represent enough detail to adequately describe existing
7-4
-------
Figure 7-2. Relationship Between Data Collection, Model Calibration,
Validation, and TMDL Procedures (adapted from McCutcheon et al., 1990)
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
conditions and generate reliable output. Finally modelers should
remember that all models have limitations and the selected model may not
be capable of simulating desired conditions. Model results must therefore
be interpreted within the limitations of their testing and their range of
application. Inadequate model calibration and [validation] can result in
spurious model results, particularly when used for absolute predictions.
Data limitations may require that model results be used only for relative
comparisons.
The basic concepts of model calibration on one synoptic data set and validation
on a second, independent synoptic data set are summarized in Figure 7-3.
7.1.2 Assessing Model Goodness of Fit
The calibration and validation process requires a series of judgements as to just
how well model performance matches observations. The question of how best to
estimate the goodness-of-fit of model output to validation data has received considerable
attention in die literature. Common practice has been to use a combination of modeler
judgment and graphical analysis to assess the adequacy of a model. However, statistical
evaluation can provide a more rigorous and less subjective approach to validation. This
section discusses several quantitative statistical techniques for assessing goodness of fit
for model predictions. Reckhow et al. (1990) provide a more detailed analysis of the
relative merits of these procedures.
Assessing Goodness of Fib t-Test
Uses: The t-test is used to compare a sample mean to a specified value. The t-
test may be used to:
(1) Compare monitoring sample mean to a model prediction;
(2) Test the magnitude of model residuals against a specified target
magnitude.
(3) Compare a distribution of modeling predictions (such as events per year
obtained from a continuous simulation, or point estimates from a Monte
Carlo simulation) to a specified or observed value.
Implementation: The null hypothesis for die t-test takes the form
x » m
7-h
-------
Figure 7-3. Model Calibration and Validation Procedure
(adapted from McCutcheon et al., 1990)
-------
The Cadmus Group> Inc.
DRAFT
June 10; 1993
where x represents the mean of the sample or model residuals and m is a fixed value.
That is, the test is set up to examine if the sample mean, or the average model residual,
both given by x, can be determined to be different from the value m at a given level of
statistical significance. It can also be thought of as testing whether the bias is
significantly different from zero, where bias is the average value of the residuals.
To test the null hypothesis we require a test statistic. The test statistic in this case
is the f-statistic, given by
x - m
t = —
*//n
where s is the sample standard deviation and n is the sample size. For use (1) above,
m would be the model prediction; for use (2), m would be the pre-specified target
absolute error of the model predictions, while for use (3), m would be a specified target
value or an observation.
The hypothesis is tested by comparison to the widely tabulated values of the t
distribution.
Assumptions and Limitations: The Mest assumes samples come from a Normal
distribution, variances are constant across distributions, and observations are
independent. The method is relatively robust to violations of the first two assumptions.
However, violations of the independence assumption can cause erroneous results.
Extensions: In some cases of model validation, the Mest may be seen as too strict
a measure. Instead of a null hypothesis that the model predicts exactly, within a given
level of statistical certainty, we might instead wish to ask whether or not the model can
predict within a given factor/of the true value. Extensions of the f-test to address this
situation are provided by Parrish and Smith (1990).
Assessing Model Goodness of Fit: Regression Analysis
Uses: A natural and intuitive method for quantitative analysis of model goodness
of fit would seem to be the formation of a linear regression between model predictions
and observations. The basic linear model is
y ¦ a + p*
where y represents model predictions and x represents calibration observations. If die
model provides a good fit, we would expect the slope of the regression, p, to be close
to 1, while the intercept, a, should be zero. We could then test whether the values of
a and p show a significant departure from these values: If the intercept is not equal to
7-8
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
zero, this indicates a consistent bias, while if J3 is not equal to 1 this indicates that model
error increases or decreases with die magnitude of the true value, suggesting inadequate
calibration. However, as warned by Reckhow et al. (1990), interpretation of the usual
regression model statistics (standard error and R2) can lead to misleading results when
applied to the model calibration problem.
Application: An ordinary least squares regression is usually formed between the
measured and calculated values. Simple Mests may then be performed to test whether
the regression slope is significantly different from 1 or the regression intercept is
significantly different from 0, indicating the presence of bias. Flavelle (1992) discusses
the range of validity of this procedure, as well as the use of regression analysis in
optimizing the calibration.
Assumptions and Limitations: The usual assumptions regarding linear
regression apply to this case. Reckhow et al. (1990) point out a number of potential
difficulties in the use of regression analysis to evaluate the behavior of environmental
models in this way:
• Often the null hypothesis will be accepted simply because the variation on
the x-variable (observations) is too small to allow meaningful hypothesis
testing.
• The regression R2 value and standard error will yield misleading measures
of the quality of fit of die model, due to lost degrees of freedom and the
unlikelihood that the true parameter values are 0,1.
• For time series observations there is likely to be autocorrelation in the
regression error. This causes bias in the evaluation of regression parameter
error. For positive autocorrelation, common in environmental applications,
the regression slope f-statistic is inflated as is the model R2.
Assessing Model Goodness of Fit: Wilcoxon Rank Sum Test
Uses: The Wilcoxon Rank Sum Test, sometimes referred to as (he Mann-Whitney
test, is a nonparametric test designed to test whether two independent random data sets
have a systematic difference. It can thus be used to compare residuals between
validation and calibration, or to compare random residuals from different model runs,
as long as the residuals have not teen constrained to zero mean in the calibration.
However, there are other potential limitations to the validity of the test, as discussed
below.
7*9
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
Application: The null hypothesis for this test states that the populations from
which the two data sets have been drawn (i.e., predictions and observations) have the
same mean, while the alternative hypothesis is that the populations have different
means.
The procedure for calculating the statistic is as follows:
1. Given two samples of size nt and n2, with m=n1+n2, combine all data and
rank from 1 to m, while preserving the identification as a member of group
1 or 2. If several data have the same value, assign them the average of the
individual ranks.
2. Sum the ranks belonging to population 1 as W.
For samples of size less than 10 (either nt or n-,) the test of the null hypothesis
must be made by comparison to critical values given by Hollander and Wolfe (1973) and
other texts on nonparametric methods. For size greater than 10, the distribution of the
statistic is approximately normal, given that the null hypothesis is true. The hypothesis
may then be evaluated from comparison to a standard normal table. When no ties are
present in the ranks, calculate
z = W-E(W)
Cvar(W)]*
where
h,KW. HtoS
12
The calculated z value is compared to a critical z value from a standard normal
table to test the hypothesis. A slight modification to the test statistic is needed to
account for the presence of tied ranks. For this, see Gilbert (1987).
Assumptions and Limitations: The primary advantage of the Wilcoxon test is
that it does not require an assumption of normality, yet retains a relatively high degree
of power. It can thus be an attractive alternative to a t-test when the normality
assumption is thought to be violated. However, as warned by Reckhow et al. (19%)/ the
Wilcoxon test can be severely effected by violations in the independence assumption,
whether due to trends or autocorrelation.
7-10
-------
The Cadmus Group, Inc. DRAFT June 10,1993
Assessing Model Goodness of Fit: Two-Sample Kolmogorov-Smirnov Test
Uses: The two-sample Kolmogorov-Smirnov test is constructed to compare two
empirical Cumulative Density Functions (CDFs), drawn from unknown distributions.
A natural application is comparing a set of model predictions to a corresponding set of
observations.
Application: Applying the two-sample Kolmogorov-Smirnov test is a simple
matter of forming empirical CDFs for the two samples (where the points on the
empirical CDF are given as i/n, for the ith ranked sample out of a total of n samples),
and evaluating D/ as the maximum value of the difference between the CDFs for each
data point. For the two-sample test, define the test statistic as (DeGroot, 1986)
nin2 ^
d = D2*
+ ty
in which case the distribution of the statistic d under the null hypothesis is the same as
that of D2, the statistic for the one-sample Kolmogorov-Smirnov test, which is tabulated
in various texts (e.g., DeGroot, 1986). As n becomes large, the critical statistic for a =
0.05 approaches 1.36/Vn. For the two-sample Kolmogorov-Smirnov test, the null
hypothesis is that the two samples come from the same distribution (i.e., that the
simulation is a good one). This hypothesis should be rejected at the a level if the
observed value of the statistic is greater than the critical value.
Assumptions and Limitations: The Kolmogorov-Smirnov test is attractive
because it is constructed directly from the observed, empirical CDFs, and makes no
assumptions regarding the underlying form of die distribution. However, the test is
based on the assumption of independence in the samples, and exact inference in the
presence of correlation would require an appropriate adjustment in the effective sample
size.
7.2 Model Accuiacy and Reliability
One important outcome of the calibration and validation process should be some
sort of estimate, either qualitative or quantitative, of the accuracy or reliability of model
predictions. This will, of course, be an important factor in deciding how to use the
model results in the estimation of the TMDL. The basic point is that models produce
only an approximation of reality. Model predictions cannot be any better than the
calibration/validation effort, and will always have some uncertainty associated with the
output. If model predictions are to be the basis of decisions, it is essential to have some
understanding of the uncertainty associated with the model prediction. For instance,
7-11
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
suppose a model for a CSO event of a given volume predicts a coliform count of 350
MPN/100 ml, well below the (hypothetical) permit requirement instream of 400
MPN/100 ml. However, the model prediction is not exact, as observation of an event
of that volume would readily show. The model must thus provide additional
information specifying how much variability to expect around the "most likely"
prediction of 350. Obviously, it makes a great deal of difference if the answer is, on the
one hand, "likely between 340 and 360", or, on the other hand, "likely between 200 and
2000".
Evaluating these issues involves the closely related concepts of model accuracy
and reliability. "Accuracy" can be defined as a measure of the agreement between the
model predictions and observations. "Reliability" is a measure of confidence in model
predictions for a specific set of conditions and for a specified confidence level. For
instance, for a simple mean estimation problem, the accuracy could be measured by the
sample standard deviation, while the reliability of the prediction (the sample mean in
this case) could be evaluated at the 95% confidence level as plus or minus approximately
two standard deviations around the mean.
An assessment of model accuracy and reliability can be an integral part of the
modeling strategy for the phased approach to TMDLs. In a phased approach, as more
data are collected and understanding of processes in the receiving waterbody increases,
it is expected that reliability or accuracy of model predictions should also increase.
During each step of the process, the TMDL developer must assess reliability of current
results to design the necessary refinements and data collection to increase accuracy.
Assessing the accuracy of individual models is also an integral part of the model
selection process. WPCF (1989) summarizes the role of model accuracy in model
selection objectives: "Modeling objectives include the need for an assessment of the
accuracy of the analysis needed for the work. This analysis will affect the requirements
for data collection, the model to be used (if any), the degree of model calibration/
validation required, and ultimately the engineering budget for the work."
Unfortunately, it is not easy to assess relative accuracy among models. The
formality and degree to which model reliability must be assessed will vary on a case by
case basis, from narrative statements to detailed quantitative analysis. It is suggested
that a quantitative analysis is usually advisable when model results are used as the
major basis for significant management decisions.
In terms of the probability of excursion of WQSs, there are two separate sources
of temporal variability to consider. These are natural variability and model uncertainty.
Natural variability concerns the variability in loading and waterbody response that
occurs as a result of precipitation sequences, and so on. Model uncertainty adds an
7-12
-------
The Cadmus Group> Inc.
DRAFT
June 10, 1993
additional layer of "noise": for instance, the simulated response to a precipitation
sequence may not be quite right. The probability of WQS excursions due to natural
variability alone can be assessed through continuous simulation over a sufficiently long
period of precipitation/flow records. However, assessment of the risk of impairment
to a waterbody should also take the accuracy of the model into account.
In the following sections we provide a brief review of techniques available to
assess the reliability, or uncertainty, associated with simulation model predictions. There
are many different techniques used to assess model reliability. This focuses on three of
the most commonly used methods: sensitivity analysis, first-order analysis, and Monte
Carlo simulation. Listed in increasing order of complexity and detail, each method is
useful for specific purposes. While sensitivity and first-order analyses are basically
analytical (rather than numerical) techniques, the nature and complexity of most wet-
weather episodic loading models and receiving water models more complex than simple
analytical methods will likely require the use of a computer to complete an analysis,
regardless of the technique chosen. Many published reports document model reliability
analysis techniques (IAEA, 1989; Cox and Baybutt, 1981; Freeze, et al., 1990; Inman and
Helton, 1988; IAEA, 1989; Marin, et aL, 1989).
7.2.1 Sensitivity Analysis
Sensitivity analysis is the least sophisticated and easiest analysis of the three to
conduct. However, this ease of use produces only rudimentary results. Consequently,
sensitivity analysis is best suited to preliminary reliability analysis and model selection
and screening.
The object of a sensitivity analysis is most clearly described by its name. This
method is used primarily to assess the sensitivity of model output to perturbations of
individual model parameters. The means of conducting such an analysis is fairly
straightforward. First, identify one, or more, parameter of interest. In most cases all
of the model parameters are chosen for the analysis. Vary each selected parameter
through its range of values, while holding all other parameters at their median, or "best-
estimate" values, and calculate the model output for each scenario. In many cases it is
sufficient to run die model with the selected parameter at only two points, its realistic
upper and lower bounds. The analysis is then repeated for each parameter identified
earlier. If the model output varies considerably for a given parameter, that parameter
is determined to have a large effect on the uncertainty in model output If the effect is
small, the model is determined to be less sensitive to the parameter.
The value of a sensitivity analysis for complex models can be improved by
conducting the effort in a formal experimental design format, in which different
7-13
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
combinations of parameters are varied on a regular, preplanned scheme. The classical
text on experimental design is Box et al. (1978).
The results of a sensitivity analysis should help to identify those parameters that
contribute most significantly to the uncertainty in the model output. It is important to
note that sensitivity analyses yield only credible ranges for model response, under
relatively strict conditions (i.e., fixing all other parameters at their median values). No
quantitative conclusions about the model accuracy can be drawn from such an analysis.
However, undertaking a sensitivity analysis can identify which parameters are likely to
provide a significant model response, and should therefore be included as a first step
in any more detailed uncertainty analysis.
7.2.2 First-Order Analysis
First-order analysis (also called variance or analytical uncertainty propagation) is
a slightly more sophisticated approach to assessing model reliability. It is used to
determine the variance of the model output as a function of the variances and
covariances of model inputs/parameters. Like sensitivity analysis, variance propagation
examines the effects of uncertainty in individual parameters on the model prediction;
however, first-order analysis produces a numerical estimate of the additional variability.
If the modeler can reasonably assume (and justify) a specific distribution on the
predicted values (e.g., a Normal distribution), then this estimated variance can be used
to compute confidence intervals for estimated values.
Depending on the nature of the model, the variance associated with one
parameter may propagate through the model very differently from the variance of
another parameter with the same level of uncertainty. That is, uncertainty in "important"
parameters will have a relatively large effect on the uncertainty associated with model
prediction; while less important variables will have a smaller impact Gearly then, die
effect of variance propagation depends on both the uncertainty associated with model
parameters and the structure of the model itself.
The object of first-order analyses is to determine the variance of the model
prediction, based on known or estimated variances of the model parameters. For linear
systems the variance of a model prediction can be derived exactly (see an introductory
statistics text for more details). As a simple example, assume we have a model of the
form
the variance of Y is then calculated from
7-14
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
nn - wy ~ niy + niy+*
wpj * p (pvrjJW5Wj\,
where
V(P) is the variance of component P„ and
p(PpPp is the correlation between P, and P;.
Substituting estimates of the parameter variances and correlations into the equation, the
modeler can then easily estimate the variance of the predicted value.
While this method does yield an estimated measure of uncertainty in the model,
it also has two important limitations. First, it does not provide a probability distribution
for the model output. Consequently, without making any assumptions about the
distribution of the model parameters or model output, only very broad probability-based
statements can be made. For example, the modeler cannot generate specific confidence
intervals for the mean value or quantiles. A second drawback of this method is that its
applicability is limited to linear or nearly linear systems. Since first-order analysis is
based on analytical techniques, we must directly compute the expression for the model
output variance. This can only be done exactly for linear systems; for nearly linear
systems the modeler can estimate the variance with a linear approximation.
7.2.3 Monte Carlo Simulation Analysis
The third method, Monte Carlo simulation, is a form of probabilistic uncertainty
analysis. The objective of this method is to build up an empirical picture of the
complete distribution function of model output over the possible range of input
parameters. For instance, if a model, denoted /, depended on a parameter, 0, with
distribution function g<0), we would wish to derive the cumulative distribution of the
model predictions over the range of 6:
CDF(f) -
Evaluation of the CDF is accomplished by a "brute force" approach, involving running
the model over and over with randomly varied parameter values and collecting the
results.
The Monte Carlo method yields not only a variance estimate but also a probability
distribution for the model prediction. This distribution is an important piece of
information, allowing the modeler to compute interval estimates and draw probability-
based conclusions about the model output.
7-15
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
To use the Monte Carlo technique, the modeler first assigns probability
distributions to each of the model parameters. These distributions should be based on
a solid combination of past experience, preliminary data screening, and expert opinion.
No inherent restrictions are placed on the form of these distributions, making Monte
Carlo analysis an easily generalizable technique.
After choosing distributions for the parameters, the modeler randomly generates
a parameter value from the appropriate distribution and inserts these values into the
model equations, yielding a predicted value. This process is repeated many (several
hundred or thousand) times, from which a sample probability distribution is generated
for the model output. This distribution reflects the overall uncertainty in the inputs to
the calculation.
The Monte Carlo technique provides several advantages over die previously
discussed approaches to reliability analysis. Most importantly, this method provides the
modeler with a probability distribution for model prediction, rather than simply an
estimate of its variance. This distribution forms the basis for computing various
estimates (e.g., mean, median, 95th percentile) and appropriate confidence intervals for
these estimates. As mentioned above, the Monte Carlo method is also applicable to a
wide variety of circumstances. For example, its use is not restricted to linear models,
wide classes of distributions may be used for input parameters, and the computations
are very straightforward. However, these advantages do not come without some cost.
Most notably, the modeler must specify distributions for the input parameters. Careful
thought must be put into assigning these distributions, as they form the basis for the
model output distribution. A frequent criticism of conclusions drawn from a Monte
Carlo simulation revolves around the choices of parameter distributions. As a result,
sensitivity to the choice of parameter distributions is an important issue to consider;
unfortunately, the effect of different distributional choices is difficult to assess. A second
potential problem lies in die computer-intensive nature of die analysis. For large,
complex models with disperse parameter distributions, Monte Carlo analysis may be
computationally infeasible. Stratified sampling techniques (e.g., Latin hypercube
sampling) may be used to reduce the effort required to obtain a representative
approximation of the CDF.
Several popular environmental fate and transport models are currently available
with Monte Carlo analysis capability (such as QUAL2E-UNCAS). Others may be
modified to perform such a function, with level of effort dependent on the clarity and
structure of die original computer code. IAEA (1989) provide a good introduction to the
details of implementing Monte Carlo methods.
7-If*
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
7.3. Interpretation of Monitoring Data and Modeling Results
Purpose: This section provides general guidance to aid interpretation and
integration of monitoring and modeling results. The information presented aims to
provide the basis for assessing receiving water impacts and assessing benefits from
implemented control and management practices, which is addressed in Section 7.4.
7.3.1 Data Accuracy, Precision, Bias, and Uncertainty
Accurate environmental assessments requires accurate information. This
information may consist of observed data, model predictions, or a combination of the
two. The first step in interpreting receiving water monitoring data and model
predictions is assessing their level of accuracy. This requires understanding that accuracy
is a measure of closeness of a measurement to the true value and it has two
components-precision and bias.
• Precision is a measure of the closeness for the collected data to their mean.
That is, do the results have a lot of "scatter?"
• Bias, most simply, is the tendency for data to be consistently and
directionally wrong. Possible examples include regular over- or under-
estimates of flow velocities or chemical concentrations.
Statistical techniques are commonly used to measure data precision through
calculation of variance, standard deviation, and confidence interval estimates for
collected data. Evaluating bias often involves using measurement standards to compare
results produced by the measuring devices used against results expected from the
standards. Such standards can include known volume flows through a calibrated
channel or known concentrations for a chemical in the standard solution analyzed. An
inexpensive analytical meter, for example, can have a low sensitivity for the variable
measured and, consequently, poor precision due to either accuracy or bias problems, or
both. But even the finest meters capable of providing the greatest precision can, when
they are poorly calibrated, provide biased results and very poor accuracy. Figure 7-4
presents a commonly used example for showing the relationship among precision,
accuracy, and Mas.
Modeling predictions similarly have characteristics of precision and bias. For
models, precision may be thought of as the "scatter" in predictions that could result from
the range of possible values of estimates of model parameters, while bias measures the
difference between model forecasts and reality. Techniques for assessing the reliability
and accuracy of model predictions were discussed in Section 7.2.
7-17
-------
Figure 7-4.
Data precision, accuracy, and bias represented by
shot patterns in targets (after Gilbert, 1987)
(a) High Bias+Low Precision *> Low Accuracy; (b) Low Bias + Low Precision ¦ Low Accuracy: (c) High
Bias+High Precision ¦ Low Accuracy; (d) Low Bias + High Precision a High Accuracy.
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
To develop confidence that data are of sufficient quality for completing an
assessment requires three practical "data screening" considerations:
(1) Are the data reasonable? This is the common-sense consideration, also
sometimes appropriately called "a laugh test." For example, some reported flow data for
a creek might be more suggestive of storm flows expected down the Mississippi River,
or some reported dissolved metal concentrations might indicate that the water might be
suitable as an ore replacement. Such results most commonly indicates an obvious error
somewhere in the analysis or data handling. While not all problems in the data are this
obvious, many can be nearly as quickly and easily spotted. For example, comparing
data to expected ranges for the variable defined as part of the quality assurance program
for the project or using some of the simple graphical data plots can aid in identifying
such problems.
(2) Are the data biased? This question can often be answered by examining the
quality assurance and quality control (QA/QC) program documentation (see Section 6.6)
and determining whether the monitoring results attain criteria established by that
program.
3) Are the data too imprecise? After concluding that the data are reasonable and
their bias is acceptable, the last step before progressing with the assessment of impact
is to determine die level of variability contained within the collected data. Statistical
estimators of means, variances, standard deviations, and confidence limits are a common
way to help assess this question. Most of the common spreadsheet and database
programs can quickly and simply complete these necessary computations.
Before proceeding with final statistical analysis of collected data, one should
examine the data summaries for those variables, particularly their means and confidence
intervals. When the random scatter (as opposed to inherent non-random variability, e.g.,
diurnal shifts in dissolved oxygen concentrations) in data collected for a variable at each
sampling location is great, the data may have a very limited value for assessing impacts.
For example, suppose mean concentrations for a constituent (e.g., lead concentrations)
appear markedly reduced in receiving waters after implementing watershed BMPs (e.g.,
increased intensity of street sweeping) compared to before BMP implementation.
Suppose also that the confidence limits from die data sets, to be contrasted for these
variables both overlap zero (or the lower detection limit for lead). Here, since the
analyzed concentrations for the individual sets of analyses are not significantly different
from zero (or the lower detection limit) individually, further statistical comparisons
between these data would not provide useful information either on differences between
data sets for the variable or on potential benefits from the BMPs. Similar analyses could
be made for data having underlying trends or cyclic variability, where the scatter can
7-19
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
be evaluated around the "true" line or curve for their inherent variability.
Overall, there is little reason to proceed with more complex analysis when
confidence intervals for sample test statistics overlap those for other sample sets that are
to be contrasted, or when confidence intervals for all contrasted data sets include zero.
Box and whisker plots is a useful graphical method for making similar comparisons of
data medians for data sets (see Wedepohl et. al., 1990, for an example application).
7.3.2 Combining Monitoring Data and Model Predictions
Monitoring and modeling are complementary processes. As previously discussed
in Section 5.1, one is not a replacement for the other and hard data are always preferable
to uncertain model predictions. However, even when monitoring data are available,
model results are also used to assess impacts and probability of water quality excursions.
At its simplest, this is because we rarely have enough monitoring data to provide a
complete picture of conditions in the receiving waterbody at all points in space and time.
Models can help fill in the gaps in the monitoring record. They can also help predict
response to conditions which have not been observed. This is particularly important for
wet-weather, episodic loads, which may depend strongly on large and infrequent
precipitation events.
The role of modeling relative to monitoring is summarized by Donigian and
Huber (1991):
Computer models allow some types of analysis, such as frequency analysis,
to be performed that could rarely be performed otherwise since periods of water
quality measurements are seldom very long. It should always be borne in mind,
however, that use of measured data is usually preferable to use of simulated data,
particularly for objectives 1 and 2 in which accurate concentration values are
needed. In general, models are not good substitutes for good field sampling
programs. On the other hand, models can sometimes be used to extend and
extrapolate measured data.
Modeling can also be viewed as a data synthesis tool. That is, the development
of a mathematical model gives model developers the ability to compile and synthesize
all, or essentially all, information known regarding the system being modeled. This
process frequently leads to identifying unknowns and specific additional monitoring
needs. In an ideal modeling application, the essence of the information contained in the
monitoring data is summarized in the model via the calibration and validation process.
The model then constitutes the most convenient summary of the responses of the
receiving waterbody.
7-20
-------
The Cadmus Group> Inc.
DRAFT
June 10,1993
The interpretation of model predictions depends in large part on the question of
whether the spatial and temporal scales of model predictions and desired results match.
For instance, using simple steady-state screening models to assess CSO impacts in
receiving waters cannot capture the actual time variability of impacts, and may thus be
inherency biased for analysis of pollutants with a short response time. However, if set
up correctly (i.e., designed to err on the side of safety), such an analysis could provide
a worst-case analysis, and thus yield information on the maximum likely impact.
More sophisticated modeling applications will generally attempt to provide
predictions at the spatial and temporal scale appropriate to the WQS. When used for
continuous (rather than event) simulation, simulation model results may be used to
predict the frequency of excursions of WQSs. This can also be accomplished by
probabilistic model applications, such as Monte Carlo simulation, in which the
simulation is made over the probability distribution of precipitation and other forcing
functions (see Section 7.2.3). In either case, model output may be analyzed for WQS
excursions, and the frequency of such excursions evaluated.
In interpreting model results, the inherent limitations of modeling must be kept
in mind. While modeling is an invaluable tool, it cannot provide exact forecasts of the
future, nor can it take the place of good observational data. Modeling of CSOs and their
impacts in receiving waters is a relatively inexact science. For instance, with sufficient
effort, TMDL developers can often obtain a fairly high degree of accuracy in modeling
the hydraulic response of a CSS. In contrast, modeling pollutant buildup/washoff,
transport in the CSS, and fate in receiving waters is considerably less exact. However,
the predictive ability of even a highly accurate hydraulic response model of a CSS is
limited because CSOs are largely the result of essentially random sequences of
precipitation events. Because of this, CSO modeling cannot produce deterministic
predictions of future events. However, it can be used to predict the expected frequency
and duration of such events.
The TMDL developer should keep in mind the various inherent limitations of
simulation modeling. The cautions expressed by Nix et al (1991) provide a useful
summary of many of these issues:
Hie limitations of all models and the frustrations of simulation are
quickly evident to the newly initiated. Sometimes the negative reaction
leads to a rejection of the whole notion of modeling. The most healthy
response, though, is one that recognizes limits and learns to treat sewer
system modeling as a science and an art Said time and time again, it bears
repeating anyway—a model is just a tool and not a replacement for sound
engineering. Models are vital to the assessment and abatement of
combined sewer overflow problems. However, any model must be placed
7-21
-------
The Cadmus Group, Inc.
DRAFT
June 10/1993
in its proper role in the overall analysis and design process, and its output
must also be interpreted with a keen awareness of inherent limitations and
assumptions....
The use of any model to simulate a combined sewer system and its
catchment is inherently limited in a number of ways. First, a computer
model cannot improve a database. It can extract information from a
database, but it cannot overcome data inadequacies. Second, no model will
produce completely accurate results because every model is incomplete
and biased in its representation of the system. Third, numbers produced
by a computer model are no more accurate than numbers produced by
hand calculations, just faster. Placing a model (like the rational method)
on a computer does not improve it. The computer model just makes the
large number of calculations tractable ....
With these caveats in mind, model interpretation should be guided by two
principles: (1) model predictions are-no better than the quality of fhe calibration, and
(2) all model predictions provide only an estimate, or best guess of future events. In
other words, model predictions are uncertain. Ideally, the levels of uncertainty present
in the model predictions should be analyzed explicitly. The predictions can then be
examined in terms of the probability of excursions of WQSs.
7.4 Evaluation of Effectiveness of Best Management Practices and Other Control
Strategies
This guidance focusses on TMDLs involving wet-weather loads. Such loads may
discharge as point or nonpoint sources, but typically involve a load component derived
from land surface runoff processes. This component of loading can be addressed
through management practices
The phased approach to TMDLs allows an iterative approach to TMDL
development, particularly where nonpoint source controls are involved. The phased
approach is defined (U.S. EPA, 1991a) "as a TMDL that includes monitoring
requirements and a schedule for re-assessing TMDL allocations to ensure attainment of
water quality standards." Where LAs for nonpoint sources are established as part of the
TMDL, the phased approach includes a description of the implementation mechanisms
and a process for the evaluation of their effectiveness. The implementation mechanisms
are typically best management practices (BMPs) for the reduction or elimination of
nonpoint source pollution. Evaluation of their effectiveness usually involves a
combination of monitoring and modeling. Because BMP evaluation plays an important
role in the estimation of TMDLs where episodic, wet-weather loads are significant, it is
7-22
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
emphasized in this section.
7.4.1 Description of Best Management Practices
A Best Management Practice (BMP) is defined in 40 CFR Part 130 as "Methods,
measures, or practices [selected by an agency] to meet its nonpoint source control needs.
BMPs include but are not limited to structural and non-structural controls and operation
and maintenance procedures. BMPs can be applied before, during, and after pollution-
producing activities to reduce or eliminate the introduction of pollutants into receiving
waters." The broadness of this definition reflects the broad category of nonpoint
pollutant sources and the still broader category of potential remedies. BMPs for
nonpoint source control are also referred to as management measures. EPA (1993)
recently has provided a comprehensive summary of BMPs for most urban and non-
urban land uses in its Management Measures guidance. The TMDL developer is
referred to this guidance for detailed information on BMPs. Management measures
described therein fall into three general categories, delivery reduction, source reduction,
and reduction of direct impacts. These have differing degrees of susceptibility to
evaluation through monitoring. The Management Measures Guidance discusses these
as follows:
Delivery Reduction. Deli very-reduction measures lend themselves to inflow-
outflow, or process, monitoring to estimate the effectiveness in reducing loads. The
simple experimental approach is to take samples of inflow and outflow at appropriate
time intervals to measure differences in the water quality between the two points. An
example is the analysis of totals suspended solids (TSS) concentrations at the inflow and
outflow of a sediment retention basin to determine the percentage of TSS removed.
Source Reduction. Source-reduction measures generally cannot be monitored
using a process design because there are usually no discrete inflow and outflow points.
The effectiveness of these measures will generally be determined by applying approaches
such as paired-watershed studies and upstream-downstream studies.
Reduction of Direct Impacts. The effectiveness of measures intended to prevent
direct impacts cannot be determined through the monitoring of loads since pollutant
loads are not generated. Instead, monitoring might include reference site approaches
where the conditions (e.g., habitat or macroinvertebrates) at the affected (or potentially
affected) area are compared over time (as management measures are implemented)
versus conditions at a representative unimpacted site or sites nearby (Ohio EPA, 1988).
This approach can be taken to the point of being a paired-watershed study if the
monitoring timing and protocols are the same at the impacted and reference sites.
7-23
-------
The Cadmus Group> Inc.
DRAFT
June 10, 1993
7.4.2 Evaluation of BMFs
The TMDL developer needs to evaluate BMPs, and other control strategies
included in a TMDL, in terms of their effect on waterbody impairment Four concepts
are integral to the discussion presented in this section that need to be defined to avoid
possible confusion (U.S. EPA, 1991). An impact is a change in the physical, chemical, or
biological quality or condition of a waterbody. An impairment is a detrimental effect on
the biological integrity of a waterbody caused by an impact that prevents obtainment of
designated uses. Biological integrity is functionally defined as the condition of the aquatic
community inhabiting unimpaired waterbodies of a specified habitat as measured by
community structure and function. And, an aquatic community is an association of
interacting populations of aquatic species within a waterbody or habitat.
There is a frequent tendency to equate any impact associated with a cultural
activity affecting a surface water as an impact resulting in impairment. However, it is
possible to have impacts that benefit aquatic communities in some systems. To help
reduce possible confusion, the terms "adverse impacts" and "beneficial effects" can be
used. Table 7-1 provides guidance on the general scope of considerations that should
be given aquatic, wetland, and riparian communities when assessing potential effects
associated with point and non-point sources. All potential adverse impacts and
beneficial effects shown in the table should be assessed, as possible, through monitoring,
modeling, and/or reference ecosystems studies. Evaluation of other special adverse
impacts and beneficial effects also may be appropriate when completing the TMDL
process.
There are three classes of techniques for quantifying the benefit of BMPs for
reducing nonpoint pollution: (1) Paired-Watershed Studies; (2) Upstream-Downstream
Studies; and (3) Use a simulation model to compute the change in pollutant
concentration and loading resulting from the installation/implementation of BMPs. Each
of the techniques has inherent advantages and disadvantages. The following sections
outline those relative advantages and disadvantages. Given the general poor reliability
of water quality modeling for pollutants, a successful strategy for BMP simulation might
involve two or even three of the above classes of efforts.
7.4.2.1 BMP Evaluation by Paired-Watershed Method
The following discussion of paired-watershed design is adapted from U.S. EPA
(1993): In the paired-watershed design there is one watershed where the level of
implementation (ideally) does not change (the control watershed) and a second
watershed where implementation occurs (the study watershed). This design has been
shown in agricultural nonpoint source studies to be the most powerful study design for
7-24
-------
The Cadmus Group, Inc.
DRAFT
June 10, 1993
Table 7-1. Guidelines useful to determine potential impacts and benefits in receiving
waters associated with discharges (expanded from Tuden et al., 1992)
Adverse impacts;
Reduced abundance of desirable species
Decreased vigor of desirable species
Loss of productivity by desirable species
Disappearance of desirable species
Increased abundance of undesirable species
Shifts in community dominance to undesirable species
Bioaccumulation of materials in tissues likely to harm human, wildlife, or
fish consumers of those tissues
Violation of downstream water quality standards, impacts to actual beneficial
uses, or degradation of groundwaters
Beneficial effects;
Improvement of water quality conditions key to ecosystem development and
preservation, e.g., pH, dissolved oxygen, ammonia
Preservation of existing, desirable riparian or aquatic spedes that would not
be maintained without discharge of the effluent
Increased diversity of desirable species
Increased productivity by desirable species
Flow augmentation
Desirable species:
Native species reflecting non-degraded conditions in the receiving water.
Species of special concern (e.g., federal and state listed "threatened" or
"endangered" species; otherwise listed "sensitive" species)
Species of special cultural interest (e.g., non-native game fish species)
Undesirable species:
Noxious spedes, not dominant under natural conditions (e.g., blue-green
algae that form blooms and cause taste or odor problems; weedy
spedes)
Non-native spedes that prey on or compete for food or habitat space with
desirable spedes (e.g., green sunfish, carp, salt cedar, zebra mussels)
7-25
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
demonstrating the effectiveness of nonpoint source control practice implementation
(Spooner et al., 1985). Paired-watershed designs have a long history of application in
forest hydrology studies. The paired-watershed design must be implemented properly,
however, to generate useful data sets. Some of the considerations to be made in
designing and implementing paired-watershed studies are described below.
In selecting watershed pairs, the watersheds should be as similar as possible in
size, shape, aspect, slope, elevation, soil type, climate, and vegetative cover (Striffler,
1965). (Further details on the selection of reference or paired watersheds is provided in
Section 7.4.3 of this guidance). The general procedure for paired-watershed studies is
to monitor the watersheds long enough to establish a statistical relationship between
them. A correlation should be found between the values of the monitored parameters
for the two watersheds. For example, the total nitrogen values in the control watershed
should be correlated with the total nitrogen values in the study watershed. A pair of
watersheds may be considered sufficiently calibrated when a parameter for the control
watershed can be used to predict the corresponding value for the study watershed (or
vice versa) within an acceptable margin of error.
It is important to note that the calibration period should cover all or the
significant portion of the range of conditions for each of the major water quality
determinants in the two watersheds. For example, the full range of hydrologic
conditions should be covered (or nearly covered) during the calibration period. This
may be problematic in areas where rainfall and snowmelt are highly variable from year
to year or in areas subject to extended wet periods or drought. Calibration during a dry
year is likely to not be adequate for establishing the relationship between the two
watersheds, particularly if subsequent years include both wet and dry periods.
Similarly, some agricultural areas of the country use long-term, multiple-crop
rotations. The calibration period should cover not only the range of hydrologic
conditions but also the range of cropping patterns that can reasonably be expected to
have an influence on the measured water quality parameters. This is not to say that the
calibration period should take 5 to 10 years, but rather that States should use careful
judgment in determining when the calibration period can be safely ended.
After calibration, the study watershed receives implementation of management
measures, and monitoring is continued in both watersheds. The effects of the
management measures are evaluated by testing for a change in the relationship between
the monitored parameters (i.e./ a change in the correlation). If treatment is working, then
there should be a greater difference over time between the treated study watershed and
the untreated (poorly managed) control watershed. Alternatively, the calibration period
could be used to establish statistical relationships between a fully treated watershed
(control watershed) and an untreated watershed (study watershed). After calibration
7-26
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
under this approach, the study watershed would be treated and monitoring continued.
The effects of the management measures would be evaluated, however, by testing for
a change in the correlation that would indicate that the two watersheds are more similar
than before treatment
It is important to use small watersheds when performing paired-watershed
studies since they are more easily managed and more likely to be uniform (Striffler,
1965). EPA recommends that paired watersheds be no larger than 5,000 acres (USEPA,
1991c).
7.4.2.2 BMP Evaluation by Upstream-Downstream Studies
The discussion of upstream-downstream studies is also taken from U.S. EPA
(1993): In the upstream-downstream design, there is one station at a point directly
upstream from the area where implementation of management measures will occur and
a second station directly downstream from that area. Upstream-downstream designs are
generally more useful for documenting the magnitude of a nonpoint source than for
documenting the effectiveness of nonpoint source control measures (Spooner et al., 1985),
but they have been used successfully for the latter. This design provides for the
opportunity to account for covariates (e.g., an upstream pollutant concentration that is
correlated with a downstream concentration of same pollutant) in statistical analyses and
is therefore the design that EPA recommends in cases where paired watersheds cannot
be established (U.S. EPA, 1991c).
Upstream-downstream designs are needed in cases where project areas are not
located in headwaters or where upstream activities that are expected to confound the
analysis of downstream data occur. For example, the effects of upstream point source
discharges, uncontrolled nonpoint source discharges, and upstream flow regulation can
be isolated with upstream-downstream designs.
It is important to note that background seasonal variation in pollutant
concentrations can compound the difficulty of discerning individual pollution sources
and BMP effects. For example, monthly mean nitrate concentration in the Fall Creek
watershed in New York varies sinusoidally in a seasonal pattern with a relative
maximum occurring each spring or late winter and a relative minimum occurring each
summer (Bouldin, 1975). A similar pattern has been observed in other watersheds
(Likens, et al., 1977; Hill, 1986). This background variation can mask or exaggerate the
effect of BMPs unless it is taken into account in designing a sampling plan.
7-27
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
7.4.2.3 BMP Evaluation with Simulation Models
A third technique for estimating the effectiveness of BMPs on wet-weather loading
is simulation modeling of the resultant loading in runoff. This is the only technique
available for direct estimation of prospective BMPs, prior to implementation. Models for
wet-weather loads, such as those discussed in Chapter 3, may be used for this purpose*
However, it is often difficult to reliably simulate the effects of small scale BMPs with
watershed-scale models. The exception would be where relatively large scale BMPs are
planned, for example the alteration of cultivation practices for the majority of an
agricultural watershed.
A full range of model types, from simple to complex can potentially be used to
simulate the effectiveness of changes in management practices in watersheds, although
simple models are too coarse in scope to simulate field scale changes. Even moderately
complex models will be poor simulators of field-scale changes if they are based on a
lumped-parameter method. The selection of appropriate simulation models for BMP
evaluation is addressed in Chapter 4.
In most cases, the final evaluation of the effectiveness of BMPs will be based on
a combination of methods. Models as the sole basis of evaluation will be applied only
in prospective evaluations, that is, in the first stage of the phased TMDL process. As the
phased approach requires monitoring to establish effectiveness of management measures,
subsequent analyses will be based on a combination of monitoring and modeling.
However, because wet-weather loading processes are driven by stochastic rainfall events,
monitoring alone will usually be insufficient to represent the full range of potential loads
that may be generated by precipitation events. Therefore, simulation of the effectiveness
of BMPs is usually necessary to complete the analysis based on monitoring results.
7.43 Establishing Appropriate Reference Conditions
Evaluating impacts to receiving water environments requires judging observed
against expected conditions in the receiving waters. This is typically done by comparing
data collected from sites where impact may occur to similarly data from the same sites
collected prior to the expected impact and to similarly collected data from one or more
appropriate reference sites. Such reference data can often provide definitive benchmark
information both for (1) establishing what environmental conditions can be attained and
(2) assessing impacts (Green, 1979; U.S. EPA, 1990). Hence, selecting reference sites that
appropriately represent reasonable obtainable target conditions for the receiving water
is key to successful assessments.
Reference sites may represent the completely unaffected state, a relatively
unaffected state, or increasing degrees of existing impact, as deemed appropriate for a
7-28
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
study. Appropriateness of a reference site should arise out of a clear understanding of
the overall environment in which the receiving water is a part. For example, some
interested parties may advocate defining unimpacted conditions as those existing in a
waterbody prior to any impacts caused by modern society. Yet no unimpacted
waterbody likely exists today according to this restrictive definition. Essentially all
waterbodies in North America are at least subtly impacted/ for example, by airborne
pollutants. But even including such impacts, conditions occurring in wilderness
waterbodies are not reasonably obtainable at present for most waterbodies nearer to and
affected more directly by modern developments. Consequently, impacts to aquatic
communities from discharges into channelized streambeds having upstream inputs from
point and nonpoint discharges cannot be directly, and perhaps not usefully, assessed
using comparisons to aquatic communities inhabiting unchannelized "wilderness"
reference sites.
Sensible assessments of receiving water conditions generally build from data
collected from appropriate reference monitoring sites. Selection of sampling sites for
assessment and reference monitoring should include the considerations on designing
monitoring programs presented in Chapter 5.
In general, the most useful reference sites are located within the receiving water
of concern, relatively near the impact-monitoring site(s), but outside of the zone of
impact associated with the pollutant source being assessed. Reference sites typically
should not include effects attributable to any significant identifiable pollutant source.
Often, appropriate reference sites within a receiving water include upstream site(s) above
downstream impact monitoring sites, or sites in unimpacted tributary streams of the
receiving water stream. Appropriate reference sites in lakes, reservoirs, or estuaries
receiving waters should be located at sites having ecologically similar conditions to the
assessment monitoring site(s), but located outside of the zone of potential impact.
As a simple example, consider the rare case where a point source discharge is the
sole source of potential impact to a receiving water. Here, an appropriate sampling
design might include at least one impact monitoring sampling site located inside the
zone of expected impacts and one reference monitoring site outside this zone. Including
more than one impact monitoring site within the zone of impact would provide data for
assessing impact variability. To evaluate the spatial extent of impacts requires spacing
sampling sites at regular, or progressively increasing, distances apart through and
beyond the zone of expected impacts.
Monitoring designs, and the process of identifying appropriate reference sites to
assess impacts from a particular source, become progressively more complex as
additional sources of potential impact to the receiving water system occur. In general,
using a site-specific reference study approach, investigators should select one or more
7-29
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
appropriate reference sites for each source of potential impact being assessed. Ideally,
each location used for site-specific reference monitoring should have environmental
conditions equal (minus impacts) to the site(s) used to monitor potential impacts for each
source of discharge being evaluated. Clearly, "sorting out" impacts caused by any
individual source under conditions having multiple impact sources can sometimes
present difficult sampling design problems. Similarly difficult problems can occur where
significant gradients for environmental variables naturally occur within the monitored
system.
Where the multiple sources of potential impacts are widely dispersed and zones
of recovery occur between zones of impact, it is often possible to distribute multiple
monitoring sites between adjacent sources of potential impact to detect possible gradients
of change (trends for either impact or recovery) associated with each source. Here, each
set of upstream sites [or those sites most distant from source(s) of impacts on other
water types—far-field sites] may provide adequa'te reference conditions for the next set
of downstream impact monitoring sites [or tike next set of sites closer to die impact
source(s)—near-field sites].
EPA presents more detailed examples of useful sampling designs for monitoring
in Klemm et al. (1990). If no other portions of the receiving waterbody exhibit these
capacities because of existing impacts, selecting reference sites in another waterbody
would be necessary. Here, investigators may select reference sites in ecologically similar
tributary streams, neighboring watershed, lakes, or estuaries, or regional reference sites.
Considerable guidance is available for using paired watersheds and "ecoregion-' reference
sites (e.g., U.S. EPA, 1990; Gallent et al., 1989; U.S. EPA SAB, 1991).
7.4.4 Evaluating Environmental Trends
Natural variations in data for water quality and other ecological variables
generally complicate abilities to assess the effectiveness of implementing source controls.
Water quality and ecological systems naturally vary over daily, seasonal, and annual
intervals. Variations in physical patterns for precipitation, wind, temperature, and solar
radiation primarily drive these changes. Random variations in these parameters over
daily intervals cause generally smaller-scale, irregular fluctuations in many ecosystem
variables. Concurrently, regular seasonal changes in these physical parameters can
produce other more rhythmic patterns of larger magnitudes.
Using mathematical or statistical methods to identify trends and segregate causes
of variability found in environmental data sets can help to define more easily changes
associated with source controls. This Section introduces concepts involved and the use
of such procedures to detect trends in monitoring data for water quality parameters.
7-30
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
These approaches can help when assessing time-dependent changes in receiving water
ecosystems associated with implementing source control alternatives.
Figure 7-5 shows a series of examples simulating how interactions among three
different time-dependent influences might affect the distribution of collected data over
time. These plots each show what could represent data from 40 samples collected at
equal time intervals over 1 Vi years. The first of these plots [Figure 7-5(a)] shows how
collected data for a parameter influenced purely by random variation might appear.
Data collected for a chemical parameter in a discharge stream could appear similar to
this, for example, as influent concentrations and treatment efficiencies vary within the
design range. Figure 7-5(b) shows how a discharge of the substance in the first plot
might affect concentrations in a receiving water that already contained seasonally
varying background concentrations of this substance.
Next, suppose there had been a gradual and progressive effort to increase the
"treatment efficiency" over the period of data collection. Then the generally flat random
distribution shown in the first of these four plots might appear as shown in Figure 7-
5(c). Here, a general underlying trend of increasing treatment efficiency could cause a
continued decrease in the measured "discharge" concentrations over time. Figure 7-5(c)
shows this with the same random pattern shown in the first plot otherwise maintained.
Then, if this substance were again "discharged," receiving water concentrations might
appear as shown in Figure 7-5(d). Obviously, if collected data were available only from
this last plot, determining natural receiving water patterns or benefits of treatment
would not likely be a straightforward process.
Patterns like those shown in the plots of Figure 7-5 could occur in monitoring data
for physical, chemical, or biological variables. But, the relationships shown in these plots
are simplified from those likely found for real monitoring data. In particular, patterns
for data collected in or downstream of source discharge likely would be complicated by
the appearance of discharge pulses. At these times, flow volumes and concentrations
of various contaminants can show often dramatic short-term increases accompanying
runoff events. Also, acutely toxic conditions that can accompany these events can
impact aquatic life in receiving waters, causing the death of resident life forms or their
migration away from the affected waters to avoid potentially toxic conditions. Either
event can cause steep short-term trends for sampled variables.
Various data-analysis techniques are available to help distinguish temporal trends
in monitoring data where additional background influences also influence numerical
values obtained for the data. When variation in collected data caused by natural
variability can be mathematically distinguished and removed, influence of trends
attributable to other causes can become increasingly clear. For example, removing
seasonal changes from environmental data could leave only background data noise to
7-31
-------
(a) Random
85
1
¦AaAA/VWV
TIME
(c) Trend + Random
Trand +
TIME
(d) Trend + Cycle + Random
TIME
Figure 7-5. Examples of compounded interactions in time series.
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
complicate assessing the benefits interpretations of control alternatives.
Detailed discussions on the selection and use of trend-analysis techniques are
presented by Gilbert (1987), Hipel (1988), Loftis et al. (1989, 1991), and other sources
cited later in this Section. Also, the final chapter of the EPA technical guidance
document entitled Monitoring Lake and Reservoir Restoration (Wedepohl et al., 1990)
presents a brief case study that introduces many considerations involved in analyses to
detect trends in total phosphorus concentration in the Neuse River, North Carolina. In
the remainder of this Section, we briefly summarize the considerations and discussion
presented with that example to select the appropriate data-analysis techniques. Similar
considerations would be necessary when selecting appropriate trend-analysis techniques
and when completing interpretations of possible trends in environmental data collected
during many aquatic environmental monitoring programs.
As background for the case example, the Neuse River drains an area of over 6,000
square miles, which includes the City of Raleigh, two upstream water supply reservoirs,
and an extensive forested area. The U.S. Geological Survey collected monthly
phosphorus data between 1981 and 1988; detected concentrations ranged from 0.13 mg/1
to 1.8 mg/1. To determine whether a trend of change in phosphorus concentrations
existed in the data collected over this period, investigators considered both parametric
models and non-parametric models (also often called distribution-free methods).
In general, several parametric methods can be used to assess trends of change.
These methods often relate the assessed changes to some physical parameter, such as
flow, depth, or retention time. But parametric methods generally include the
assumptions that linear relationships exist between the analyzed variables and that data
for these variables have normal, bell-shaped distributions around their means.
Sometimes, data transformations can be used to achieve distributions reasonably
resembling normal (McLeod et al., 1983). Other transformations can be used to
approximate linear relationships between variable data sets. Similarly, most parametric
methods also require that the data not include seasonal influences or not be
autocorrelated, i.e., the data are not correlated in time with any prior and subsequently
collected data for the variable. Other methods are available to transform data to remove
seasonal influences (Taylor and Loftis, 1989).
Where the assumptions of parametric methods can be reasonably met and where
the trend is believed to be continuous, instead of abrupt or stepwise, it is often
appropriate to use ordinary least square regression to determine the possible slope for
the trend of change. This approach could be used, for example, to assess the trend
shown by the data in Figure 7-5(c).
When evaluating possible effects that a single event (e.g., a pulse discharge)
7-33
-------
The Cadmus Group; Inc.
DRAFT
June 10,1993
produces in a receiving water, the Student's f-test can sometimes be used, if assumptions
underlying this test are met. This test allows comparing data collected prior and after
the event to evaluate possible changes. However, more sophisticated parametric trend
analysis techniques are more appropriate when the normally distributed data include
seasonal influences or when they are autocorrelated. For additional information about
using parametric trend analysis techniques, including Box-Jenkins models, the texts of
Gilbert (1987) and Pankratz (1983), and articles by Montgomery and Reckhow (1984) and
Berryman et al. (1988) can be consulted.
In the analysis for the Neuse River example, the data failed to meet the
requirements for using parametric trend analysis methods: the data were significantly
skewed and they displayed a seasonal cycle. 61 fact, parametric techniques often are not
appropriate and should always be cautiously applied to environmental data, since many
water quality and other environmental parameters commonly have non-normal and
seasonal distributions. Therefore, the Neuse River investigators opted for non-
parametric methods to assess trends in their data. While non-parametric methods
generally are less powerful than parametric methods, they do not require that the data
have a generally normal distribution about their mean. But most non-parametric
methods still require that the analyzed data contain neither seasonal or autocorrelated
relationships.
The Neuse River data were analyzed using the seasonal Kendall's Tau Test. This
test was selected also because it can assess seasonally varying data and it is not overly
sensitive to extreme values (Gilbert, 1987; Loftis et al., 1989). These are common
problems in many environmental data sets. Following data preparation and
autocorrelation testing, as discussed by Wedepohl et al. (1990), this test revealed no
significant trend of change in phosphorus concentrations over the seven years of
monitoring. Consequently, this study indicates that (1) additional measures to control
phosphorus entering the Neuse River and (2) additional monitoring of this river are
warranted.
When analyzing water quality and ecological data for trends, we often want to
assess possible simultaneous trends in several variables at several monitoring stations.
Most trend analysis procedures, including those discussed above (excluding the seasonal
Kendall's method)/ are univariate procedures, i.e., they are able to assess trends in single
variables at single locations. These techniques are inappropriate for multivariate or
multiple location analyses. Additional multivariate tests, which are generally less widely
known and used, are available to assess such data. These methods were recently
reviewed by Loftis et al. (1991) to assess their performance with both serially
independent and serially correlated data. They report the strengths and limitations of
the five parametric and nonparametric procedures evaluated.
7-34
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
7.5 Assessing Water Quality Impacts on Fish and Other Aquatic Life
To assess the protection of beneficial uses of a waterbody from a proposed control
strategy it is often necessary to assess or predict water quality impacts on aquatic life.
This brief overview introduces several common criteria and discusses some of the
possible consequences that altered water quality has on aquatic life. Water quality
variables discussed include temperature/ pH, nutrients, dissolved oxygen, potentially
toxic chemicals, and suspended sediment. This discussion draws heavily from U.S.
EPA's 1986 Quality Criteria for Water, on an earlier review completed by the American
Fisheries Society (Thurston et al., 1979), and a review for the USDA Forest Service
(Marcus et al., 1990).
7.5.1 Temperature
Most biological and chemical processes in aquatic environments ultimately are
regulated by water temperature. Fish and essentially all other aquatic animals are cold
blooded (poikilotherms); thus, their metabolism, reproduction, development, and scope
for activity is largely controlled by environmental temperatures. Similarly, aquatic plant
photosynthesis and respiration, chemical reaction rates, gas solubilities, and microbial
mediated processes including decomposition and nutrient cycling are also temperature
dependent. In fact, the Federal Water Pollution Control Administration in 1967
described temperature as "a catalyst, a depressant, an activator, a restrictor, a stimulator,
a controller, a killer, one of the most important and influential water quality
characteristics to life in water" (U.S. EPA, 1986). Section 5.6 introduced the role of
thermal characteristic in the dynamics of most receiving water environments.
The present criterion to protect freshwater aquatic life is based on "the important
sensitive species" resident during the time of concern and consists of two upper
temperature limits (U.S. EPA, 1986), The first limit is based on short (i.e., over durations
of minutes) exposure, is computed using an equation presented in the EPA criterion, and
uses data presented in a National Academy of Sciences document The second limit is
based on a weekly maximum average temperature, which changes with season, with
reproductive stage present, to maintain species diversities, or to prevent nuisance
growths of organisms. For rainbow and brook trout adults and juveniles, the maximum
weekly average temperature for growth during the summer is listed as 19°C, and the
short-term maximum temperature limit for survival during summer is 24°C (U.S. EPA,
1986). This report also lists 9°C as the average weekly maximum temperature reported
for spawning by these species, and 13°C as the short-term maximum reported for
survival of their embryos. These present temperature criterion, however, present
numerous interpretational problems with respect to defining "important sensitive
species" and "short-term" (Thurston et al., 1979).
7-35
-------
The Cadmus Group; Inc.
DRAFT
June 10,1993
Thermal criteria for marine environment are much more straightforward. To
protect the characteristic of indigenous marine communities from adverse thermal
effects, U.S. EPA (1986) maintains that
a. The maximum acceptable increase in the weekly average
temperature resulting from artificial sources is 1°C (1.8°F) during all
seasons of the year, providing the summer maxima are not
exceeded; and
b. Daily temperature cycles characteristic of the waterbody segment
should not be altered in either amplitude or frequency.
7.5.2 pH
In natural waters, pH is primarily regulated by the solution of atmospheric carbon
dioxide, which reacts with water to form carbonic acid and then disassociates to
hydrogen and bicarbonate ions. Distilled water at equilibrium with atmospheric COa has
a pH of 5.6 at sea level. But natural waters contain various dissolved salts and organic
chemicals derived from watershed rocks, soils, and organisms that tend to buffer the
natural acidities and raise the pH of surface waters. Over approximately the past 15
years it also has become increasing apparent that the pH of weakly buffered aquatic are
often influenced by the deposition of atmospheric adds (e.g., Baker et al., 1990).
As with temperature, the concentration of hydrogen ions is an important regulator
of many chemical and biological processes in aquatic and marine environments. (Most
accurately, it is the chemical activity of hydrogen ions, rather than their concentration,
that is reflected by measured pH levels.) For example, pH primarily defines the
chemical natures of dissolved ions in waters, the directions of chemical reactions, the
adsorption of chemicals onto organic and inorganic particles, plus the availability and
toxicity of chemical to organisms. Similarly, uptake and release rates for ions across
gills, die primary method of ion regulation for aquatic animals, is at least partly pH
dependent. Environmental conditions beyond their natural pH limits can produce stress
and cause mortality of organisms.
The criteria range to protect freshwater aquatic life is pH 6.5 to 9.0 and 6.5 to 8.0
for marine life (U.S. EPA, 1966). Many fish and other aquatic life, however, are well able
to survive and reproduce at pH levels outside of this range. Also, not all species and
not all life stages of most species are equally sensitive to pH changes. For example, as
acidity levels increase, brook trout are generally less sensitive than brown trout, which
are in turn less sensitive than rainbow trout hatching and larval stages are the life stages
for these spedes that are most sensitive to acidify (Marcus et al., 1986). Brook trout
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
populations frequently have been found to inhabit waters have pH levels less than pH
5.5 (e.g., Schofield and Trojnar, 1980).
7.5.3 Nutrients
The two nutrients of greatest potential concern in aquatic systems are nitrate and
phosphate. These nutrients are the two most related to the eutrophication of surface
waters, the associated nuisance growths of algae, and the development of other noxious
conditions. The problem of excess nutrient enrichment and eutrophication of surface
waters of all types is a long recognized problem, dating bade to at least 1907
(Hutchinson, 1969,1973). According the National Academy of Science (NAS, 1969: pages
3-4),
"The term 'eutrophic7 means well-nourished; thus, 'eutrophication' refers
to natural or artificial addition of nutrients to bodies of water and to effects
of added nutrients. Eutrophication of lakes is a natural process that can
be greatly accelerated by man. Eutrophication is an aspect of aging; it
increases the rate at which lakes disappear. Some disagreement exist as to
the applicability of the term to other bodies of water. Streams do not age
in the same sense as lakes, although added nutrients will increased their
productivity. .... When the effects are undesirable, eutrophication may
be considered a form of pollution."
The effects of nutrient enrichment differ between streams and lakes. In streams,
the downstream flow of nutrients out of the immediate system typically is more
important. Further, the development of excess algal biomass is often reset by removal
of attached algae during flood scour events (Lohman et al., 1992). The question of
whether the term eutrophication is appropriate for streams was particularly brought to
focus by Hynes (1969: page 188) in that NAS volume:
"The term 'eutrophic/ which was coined for application to lakes, has
acquired so many connotations of aging and evolution of the environment
that it cannot properly be applied to running water. Unlike a lake, a
stream has no allotted life-span; it is an ongoing phenomenon, and if it
ages, it does so only in die sense of erosion toward base level."
Thus, Hynes suggested that, with respect to streams, it may be better to just use
"nutrient enrichment" rather than "eutrophication" as a term. He also acknowledged that
problems related to discharges of excess nutrients to either lakes or streams do, in fact,
exhibit many commonalities. While there has been some disagreement regarding the
application of the term eutrophication to flowing waters, it has continued to be commonly
7-37
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
used to describe the processes and problem of nutrient enrichment in streams and rivers.
Certainly, it should also be recognized that not all forms of eutrophication or
nutrient enrichment necessarily produce undesirable consequences. For example,
considerable effort has been directed at learning approaches for maintaining correct
nitrogen and phosphorus balances to apply under controlled conditions to enhance
salmon production in nutrient-poor lakes along coastal western Canada (e.g.,
Barraclough and Robertson, 1971; Stockner, 1981). In fact, Hynes (1969: page 194) ended
his consideration of nutrient enrichment of streams with,
"In summary, then, we can say that enrichment produces fairly obvious
effects on small watercourses and that in moderation these may be
beneficial from our point of view, since they increase production of such
things as game fish. Greater amounts produce definitely deleterious
effects, although up to a point this may not be true in countries where the
value of fish is reckoned in weight of protein rather than quality. In larger
rivers the effect is uncertain, but probably bad, and we may make it worse
by constructing impoundments. It should, however, be emphasized here
that in contrast to a lake, which can be made eutrophic and then probably
remains in that state, a stream or river has to be continuously enriched.
It can, therefore, be rescued and restored."
What kinds of adverse impacts accompany excess nutrient loading and
eutrophication problems? Often, one of the first responses is by phytoplankton (free-
floating microscopic algae) in standing waters and periphyton in flowing waters.
Periphyton is the assemblage of organisms that grow on underwater surfaces and is
commonly predominated by algae, but also can include bacteria, yeasts, molds, protozoa,
and other colony forming organisms. According to EPA's manual Biological Field and
Laboratory Methods for Measuring the Quality of Surface Waters and Effluents (Weber, 1973),
"Excessive growth stimulated by increased nutrients can result in large,
filamentous streamers [and mats] that are aesthetically unpleasing and
interfere with such water uses as swimming, wading, fishing, and boating,
and can also affect the quality of the overlying water. Photosynthesis and
respiration can affect alkalinity and dissolved oxygen concentrations of
lakes and streams. Metabolic byproducts released to the overlying water
may impart tastes and odors to drinking waters drawn from the stream or
lake, a widespread problem throughout the United States. Large clumps
of growth may break from the site of attachment and eventually settle to
form accumulations of decomposing, organic, sludge-like materials."
There is no clear-cut definition of exactly what constitutes nuisance levels of
7-38
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
periphytic algae. However, it has been suggested that nuisance biomass of filamentous
periphytic algae may be represented by a level greater than 100-150 mg-chlorophyll-a/m2
(Horner et al., 1983; Welch et alv 1988).
What are the concentrations of nutrients that can produce problems? No criterion
is provided by the EPA for either of these two nutrients with respect to the control of
eutrophication. U.S. EPA's (1986) criteria document does discuss the toxic potential for
nitrates to fish. This report concludes that nitrate-nitrogen concentrations at or below
90 mg/1 should be protective for warmwater fishes, while concentrations at or below
0.06 mg/1 should be protective for salmonid fish. This guideline for salmonids is based
on very limited data, and many natural salmonid waters have nitrate concentrations
exceeding this level.
U.S. EPA (1986) also suggests as a guideline to prevent nuisance algal growths
and limit cultural eutrophication that total phosphates as phosphorus should not exceed
0.1 mg/1 in any stream or other flowing water, exceed 0.05 mg/1 in any stream at the
point where it enters a lake or reservoir, or exceed 0.025 mg/1 in any lake or reservoir.
Golterman (1975) suggests that, in general, eutrophication may occur in surface
waters that have nitrate-nitrogen concentrations above 0.3 mg/1 and phosphate-
phosphorus concentrations above 0.02 mg/1. Similarly, Wetzel (1975) reports that
eutrophication in lakes can begin as total inorganic nitrogen concentrations exceed 0.5
mg/1 N, and serious hypereutrophication problems occur at concentrations exceeding
1.5 mg/1 N. He also reports that eutrophication in lakes generally occur at total
phosphorus concentrations exceeding 0.03 mg/1 P, and hypereutrophication generally
occurs at concentrations greater than 0.1 mg/1.
These concentrations for lakes are very similar to those long associated with
problems in flowing waters. For example, Mackenthun (1969: page 41) summarizing
Muller's 1953 review of nutrients in flowing water systems, noted "... that excessive
growths of plants and algae in polluted waters can be avoided if the concentration of
nitrate nitrogen is kept below about 0.3 mg/1 and the concentration of total nitrogen is
not allow to rise much above 0.6 mg/1." Mackenthun (1969: page 41) further concluded,
"A considerable judgement suggests that to prevent biological nuisances, total
phosphorus should not exceed 100 p.g/1 P [= 0.1 mg/1 P] at any point within the flowing
stream, nor should 50 jig/1 be exceeded where waters enter a lake, reservoir, or other
standing waterbody. Those waters now containing less phosphorus should not be
degraded."
Other experiments in phosphorus-limited flowing systems suggests that very low
soluble reactive phosphorus (SRP) concentrations may be required to avoid periphytic
biomass at nuisance levels. For example, less than 0.001 mg/1 SRP was recommended
7-39
-------
The Cadmus Group> Inc.
DRAFT
June 10,1993
for the Spokane River (Welch et al., 1989), and less than 0.025 mg/1 SRP from
experiments in laboratory channels (Horner et al., 1983).
Analyses of various plant tissues (including algae) indicate that total nitrogen and
total phosphorus concentrations occur within plants in the atomic ratio of 15:1 to 16:1
(Stumm and Morgan, 1970). Also, because of biological nutrient uptake by plants, the
global average of TN:TP in surface waters is also in this same range. This ratio is
generally considered the optimal ratio for general plant growth and is commonly cited
as an important indicator of relative nutrient limitations by these two nutrients Higher
ratios indicate possible phosphorus limitation, while lower ratios indicate possible
nitrogen limitation.
A TN:TP ratio of 10:1 by weight (e.g., mg/1) in aquatic ecosystems was found to
be the ratio below which nitrogen fixation by blue green algae is common (Fleet et al.,
1980). The scientific literature further indicates that at TN:TP ratios of less than 29:1 by
weight, blue-green algae maintain significant populations (>10%) within algal
communities and blooms by blue-green algae are exceedingly, common (Smith, 1983).
Part of the reason for this is that blue green algae are generally more efficient than many
other taxonomic groups of algae in acquiring nitrogen but less efficient at getting
phosphorus from surface water environments. At TN:TP ratios of greater than 29:1 blue
green algae generally comprise less that 10% of algal communities. Knowledge of these
ratios for surface waters can help to target potential sources of nutrients and to guide
corrective management actions aimed at reducing possible eutrophication problems.
It also is sometime useful to recognize that every gram of P in the water can grow
115 grams of algal biomass dry weight or about 500 grams wet weight, assuming algae
is about 80 percent water. This estimate is based on die average chemical composition
of algae (cf, Stumm and Morgan, 1970).
In conclusion, it is clear that uncontrolled and excessive additions of nutrients to
surface waters can and has severely damaged receiving water ecosystems. Excessive
algal growth in can cause diurnal depletion of dissolved oxygen to unacceptable levels.
The question often becomes one of determining the nutrient concentrations or loads
beyond which the ability of the receiving water to assimilate the inputs is overloaded.
At what concentration is the limits of the system's natural function exceeded, causing
adverse impacts to begin? Determining these loading rates are often a primary objective
of the TMDL process. Without a TMDL study, expert opinion backed by scientific
literature must be used to project nutrient concentrations that will cause potentially
adverse impacts.
7-40
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
7.5.4 Dissolved Oxygen
Determining dissolved oxygen concentrations in surface waters is very informative
because its concentration reflects the integrated health of the aquatic community. In fact,
maintenance of adequate concentrations of dissolved oxygen in receiving waters is one
of the key indicators that the discharge of nutrients and other chemicals causing
BOD/COD is within the assimilative capacity of the receiving water. EPA maintains
water quality criteria for dissolved oxygen only in freshwaters (Table 7-2). In general,
any time dissolved oxygen concentrations dips below 4.5 mg/1 in cold waters or below
3.5 mg/1 in warm waters, there is reason for concern that the chemical and biological
integrity of the system is threatened. Site-specific studies would be necessary, however,
to appropriately evaluated the degree of this magnitude and degree of this threat relative
to the species in inhabiting and the obtainable use defined for the water.
Table 7-2. Water quality criteria for ambient dissolved oxygen concentrations
(U.S. EPA, 1986)
Coldwater Criteria
Warmwater Criteria
Time
Early Life
Stagew
Other Live
Stage
Early Life
Stage?
Other Live
Stage
30-Day Mean
NA3
6.5
NA
5.5
7-Day Mean
9.5 (6.5)
NA
6.0
NA
7-Day Mean
Minimum
NA
5.0
NA
4.0
1-Day
Minimum4,5
8.0 (5.0)
4.0
5.0
3.0
1 These are water column concentrations recommended to achieve the required intergravel
dissolved oxygen concentrations shown in parentheses. The 3 mg/1 differential is discussed in
the criteria document For species that have early life stage exposed directly to the water
column, the figures in parentheses apply.
2 Includes all embryonic and larval stages and all juvenile forms to 30-days following hatch.
3 NA « not applicable
4 For highly manipulatable discharges, further restrictions may apply.
5 All minima should be considered as instantaneous concentrations to be achieved at all times.
7-41
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
7.5.5 Toxic Chemicals
Aquatic life can be affected by an increasing diversity of potentially toxic
chemicals. A review of the toxicity and potential impacts of all of these chemicals is
beyond the scope of this report. But, Table 7-3 lists the lowest observed effects levels
obtained from short-term acute toxicity and long-term chronic toxicity tests for a variety
of toxicants potentially encountered in receiving waters; water concentrations less than
the indicated levels may not affect aquatic life. The information provide a guideline to
determine potentially hazardous conditions for aquatic life. For a specific water, actual
toxic concentrations are often likely to be either greater or lesser than the reported
values. For example, the toxicity of ammonia increases as either temperature or pH of
the water increases (Thurston et al., 1979).
As another example, the toxicity of many metals and other chemicals is affected
by hardness. That is, actual instream toxicities of metals to aquatic organisms does not
depend solely on the total concentrations of the metals in the water. A variety of
studies, show that water quality characteristics, especially hardness, alkalinity, pH, and
chelating organic materials, can affect the aquatic chemistries, toxicities, and
bioavailabilities of metals (Black et al., 1975; Parkhurst et al., 1984). Studies also reveal
that not only can calcium hardness, in particular, affect the instream chemistry of many
metals, but it can affect the physiological susceptibilities of the organisms to the potential
toxicities of many metals (Davies and Woodling, 1980; Mount et al., 1988; Parkhurst et
al., 1984; Pascoe et al., 1986).
In addition, potentially toxic chemicals are rarely present singularly; toxicities
derived from multiple chemical sources may be additive, subtractive, or multiplicative.
Therefore, when evaluating the potential toxicities to fish by chemicals in surface waters,
evaluations must include careful determination of and potential interactions with other
chemicals present.
7.5.6 Suspended Sediment
Turbidity is a measure of the scattering and absorption of light by dissolved and
particulate matter in water (Lloyd, 1987). Usually, turbidity and suspended sediment
concentration are highly correlated; thus, turbidity can provide an index of suspended
sediment concentrations (Lloyd et al., 1987). Because murky water absorbs more heat
than clear waters, increased suspended sediment loads can cause water temperatures to
increase (Hynes, 1970). Water with a temperature of 5°C is able to carry 2 to 3 times
more sediment than 27°C waters (Heede, 1980).
7-42
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
Table 7-3. 1986 Quality Criteria for Water (U.S. EPA, 1986).
Chemical
Freshwater Concentration
(pg/1)
Marine concentration (jig/1)
Acute
Criteria
Chronic
Criteria
Acute
Criteria
Chronic
Criteria
Aldrin
3.0
1.3
Ammonia
Criteria are pH and temperature dependent
Antimony
9,000*
1,600*
Benzene
5,300*
5,100*
700*
Beryllium
1301
5.3*
Cadmium
3.92
1.1*
43
93
Chlordane
2.4
0.0043
0.09
0.004
Chlorine
19
11
13
7.5
Chloroform
28,900
1,240
Chromium (Hex)
16
11
1,100
50
Chromium (Tri)
1,70a1
2102
10,300*
Copper
18*
12*
2.9
2.9
Cyanide
22
5.2
1
1
DDT
1.1
0.0010
0.13
.001
DDT Metabolite
(DDE)
1,050*
14*
DDT Metabolite
(TDE)
0.6*
3.6*
Dieldrin
2.5
0.0019
0.71
0.0019
Endrin
0.18
0.0023
0.037
0.0023
Iron
1,000
7-43
-------
The Cadmus Group; Inc.
DRAFT
June 10, 1993
Table 7-3. 1986 Quality Criteria for Water (U.S. EPA, 1986).
Chemical
Freshwater Concentration
(Hg/D
Marine concentration (]ig/l)
Acute
Criteria
Chronic
Criteria
Acute
Criteria
Chronic
Criteria
Lead
822
3.22
140
5.6
Malathion
0.1
0.1
Mercury
2.4
0.012
2.1
0.025
Nickel
1,800*
962
75
8.3
Parathion
0.04
PH
6.5-9.0
6.5-8.5
Phenol
10,200*
2,560*
5,800
Selenium
260
35
410
54
Silver
4.12
0.12
2.3
Hydrogen Sulfide
2
2
Thallium
1,400*
401
2,130'
Toluene
17,500*
6,300*
5,000*
Zinc
32 O2
47
95
58
1 Insufficient data to develop criteria. Value presented is the Lowest
Observed Effects Level (LOEL).
Hardness dependent criteria (100 mg/1 hardness used for reported value).
The EPA has narrative criteria for suspended solids and turbidity:
Settleable and suspended solids should not reduce the depth of the
compensation point for photosynthetic activity by more than 10 percent for
the seasonally established norm for aquatic life.
7-44
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
Table 7-3. 1986 Quality Criteria for Water (U.S. EPA, 1986).
Chemical
Freshwater Concentration
(pg/D
Marine concentration (pg/1)
Acute
Criteria
Chronic
Criteria
Acute
Criteria
Chronic
Criteria
Lead
00
CM
to
140
5.6
Malathion
0.1
0.1
Mercury
2.4
0.012
2.1
0.025
Nickel
1,800*
96*
75
8.3
Parathion
0.04
PH
6.5-9.0
6.5-8.5
Phenol
10,200'
2,560'
5,800
Selenium
260
35
410
54
Silver
4.12
0.12
2.3
Hydrogen Sulfide
2
2
Thallium
1,400'
40'
2,130'
Toluene
17,500*
6,300'
5,000'
Zinc
320*
47
95
58
Insufficient data to develop criteria. Value presented is the Lowest
Observed Effects Level (LOEL).
Hardness dependent criteria (100 mg/1 hardness used for reported value).
"""""XsanoffierexampenKeTo^
by hardness. That is, actual instream toxicities of metals to aquatic organisms does not
depend solely on the total concentrations of the metals in the water. A variety of
studies, show that water quality characteristics, especially hardness, alkalinity, pH, and
chelating organic materials, can affect the aquatic chemistries, toxicities, and
bioavailabilities of metals (Black et al., 1975; Parkhurst et aL, 1984). Studies also reveal
that not only can calcium hardness, in particular, affect the instream chemistry of many
metals, but it can affect the physiological susceptibilities of the organisms to Hie potential
7-44
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
toxicities of many metals (Davies and Woodling, 1980; Mount et al., 1988; Parkhurst et
al., 1984; Pascoe et al., 1986).
In addition, potentially toxic chemicals are rarely present singularly; toxicities
derived from multiple chemical sources may be additive, subtractive, or multiplicative.
Therefore, when evaluating the potential toxicities to fish by chemicals in surface waters,
evaluations must include careful determination of and potential interactions with other
chemicals present.
7.3.5.6 Suspended Sediment
Turbidity is a measure of the scattering and absorption of light by dissolved and
particulate matter in water (Lloyd, 1987). Usually, turbidity and suspended sediment
concentration are highly correlated; thus, turbidity can provide an index of suspended
sediment concentrations (Lloyd et al., 1987). Because murky water absorbs more heat
than clear waters, increased suspended sediment loads can cause water temperatures to
increase (Hynes, 1970). Water with a temperature of 5°C is able to carry 2 to 3 times
more sediment than 27°C waters (Heede, 1980).
The EPA has narrative criteria for suspended solids and turbidity:
Settleable and suspended solids should not reduce the depth of the
compensation point for photosynthetic activity by more than 10 percent for
the seasonally established norm for aquatic life.
This criterion is based on the depth in the water column at which planktonic
photosynthesis equals respiration (i.e., the compensation point), and it provides a generally
weak bases for extrapolating possible effects in other groups or in flowing water
environments.
The European Inland Fisheries Advisory Commission concluded that suspended
sediment can affect aquatic organisms by killing them directly/ by reducing growth rates
and resistance to disease, by preventing successful development of eggs and larvae, by
modifying natural movement or migration patterns, or. by reducing the natural
availabilities of food (U.S. EPA, 1986). A review completed by the National Academy
of Science suggested that a limit of 25 mg/1 of suspended sediment would provide high,
80 mg/1 moderate, 400 mg/1 low, and over 400 mg/1 very low levels of protection for
aquatic organisms (Thurston et al., 1979). Lloyd (1987) suggested, based on his review
of turbidity studies in Alaska, that a water quality standard that permitted an increase
of 25 NTUs (nephelometric turbidity units) above ambient would provide moderate
7-45
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
protection for dear, coldwater stream habitats.
The effects of fine sediment on fish, particularly trout and related species, in
streams has received particular research attention (see Chapman, 1988, and Marcus et
al., 1990, for extensive reviews). We present a brief overview of some of the results in
the balance of this section.
Though researchers do not agree on the exact size definition of fine sediment, it
is generally taken to be less than 6.3 mm in diameter (Chapman, 1988). As discussed
in Section 5.6, sediment transport in streams is a very complex relationship involving at
least 30 variables (Heede, 1980). But, in general, transport of fine sediment may be via
saltation along the stream bottom or suspension in the water column, with discharge and
channel slope proportional to the quantity and size of transported sediment (Hasfurther,
1985). Typically, transport of sediment is greater on the ascending limbs of storm
hydrographs, but this is due more to the supply of sediment rather than the hydraulics
of the flows (Sidle, 1988; Sidle and Campbell, 1985).
Everest et al. (1987) and Chapman (1988) suggested that limited fine sediment
may be beneficial to some salmonids by contributing to increased invertebrate
productivities, and that the adverse consequences of fine sediment introduction to trout
streams have been sometimes has been overstated. Nonetheless, the transport and
deposition of fine sediment deleteriously affect survival throughout the life history of
salmonids is many stream systems.
Suspended sediment may directly or indirectly influence the survival of aquatic
organisms directly by clogging and damaging respiratory organs. Concentrations greater
than 100 mg/1 have reduced survival of juvenile rainbow trout (Herbert and Merkens,
1961). Reductions in growth or feeding of salmonids were associated with turbidity over
25 nephelometric turbidity units (NTU) (Olson et al., 1973; Sigler et al, 1984; Sykora et
al., 1972). Since salmonids are considered to be sight-feeders, the reduction in light
transmission caused by high turbidity may result in less feeding and decreased growth
(Berg, 1982). In response to turbidity, salmonids may change their use of cover or
reduce territoriality (Berg andNorthcote, 1985). When given the opportunity, juvenile
coho salmon avoided turbid water (Bisson and Bilby, 1982). Despite these impacts,
salmonids often successfully inhabit streams with seasonally high turbidities, perhaps
due to behavioral modifications and to limited exposure to concentrated suspended
sediments.
Deposition of fine sediment can acutely affect survival of salmonids (1) during
intragravel incubation of eggs and alevins; (2) as fingerlings; and (3) throughout winter
(Chapman and McLeod, 1987). Timing, source, and quantity of deposited sediment can
affect survival. Increasing proportions of fine sediment in substrates have been
7-46
-------
The Cadmus Group> Inc.
DRAFT
June 10,1993
associated with reduced intragravel survival of embryonic trout, char, and salmon
species. Increases in fine sediment can directly limit survival-to-emergence only by
entrapping alevins. The potentially greater influence on survival by increased sediment
deposition is the decrease in dissolved oxygen concentration coupled with reduced
intragravel water flow (Chapman, 1988). However, most studies evaluating the impacts
of fine sediment on embryonic survival have been conducted in the laboratory; few or
no field studies have satisfactorily quantified actual impacts (Chapman, 1988).
Fingerling density has often been associated with low concentrations of fine
sediment deposited between and on the surface of larger substrate particles, i.e.,
embeddedness (Burns and Edwards, 1985). Chapman and McLeod (1987) reported that
the relation between the rearing densities of salmonids and fine sediment was equivocal.
Instead, they suggest that changes in stream morphology caused by fine sediment may
outweigh the effects of embeddedness on fingerling survival.
Declining water temperatures in winter may cause salmonids to seek refuge
within the intersticial spaces of the substrate. Deposition of fine sediment could also
restrict winter cover for adult fish by filling in low velocity habitats, e.g. pools, and
undercut banks (Bjornn et al., 1977).
7.6 Interpretation of Biological Data
This section provides general guidance and additional reference sources with
information that can aid interpretation of monitoring data for aquatic biological
communities. The discussion emphasizes impact assessment. It begins with simple
evaluation steps to qualitatively evaluated sample data and introduces methods and
cautions for conducting quantitative community analysis. The final three subsection
reviews additional sources for specific guidance for assessing algae, benthic
macroinvertebrates, and fish.
7.6.1 Qualitative Evaluation of Sample Data
After the investigator is satisfied that an adequate monitoring plan was used and
the data are of adequate quality (often as defined by the QA/QC protocol), inspection
and interpretation of the data can begins. A common and useful first step is to examine
the total number of species present (species richness), note differences in taxonomic
compositions among sampling times and sampling sites, and review the environmental
requirements for those species most commonly found in the samples. Various
documents, which we note later in this section, provide considerable useful information
on the sensitivities and tolerances for many spedes to various environmental conditions.
Some of these documents also list those species that are widely recognized as key
7-47
-------
The Cadmus Group, Inc.
DRAFT
June 10, 1993
indicator species that have particular sensitivities or tolerances to individual environmental
variables. The taxa lists developed from the samples should be scanned to determine
whether any of these indicator species occur. When the results from these comparisons
are compiled and this information is qualitatively integrated with the physical and
chemical monitoring results, a useful "first-cut" qualitative evaluation of the monitored
environment often results. (See below for two important cautions on this approach.)
Other key qualitative information to look for when first examining the biological
sample data is obvious evidence of possible seasonal cycles in the appearances and
disappearances of individual taxa in the collected data. This is especially valuable when
examining algae and invertebrate samples results. Are seasonal patterns repeated year
to year? Do maximum and minimum densities (e.g., numbers per sample) or biomasses
remain relatively constant each year? Is a trend of increase or decrease suggested by the
data? While seasonal patterns and year-to-year differences are also important when
examining data on fish, it is also important to determine whether some year classes
(individuals recruited to the population from a single year of spawning) may be missing.
Severe environmental stress can cause spawning complete spawning failures in some
years, producing gaps in the year-class/age-structure of some fish populations. Often
these qualitative evaluations of the collected data are useful in helping to focus, or
sometimes to redirect, subsequent statistical analysis of the collected data.
7.6.2 Quantitative Community Analysis
Many community indices are available to aid in die quantitative summarization
and interpretation of collected data. Documents noted at the end of this section provide
considerable helpful guidance on use and interpretation of many of these indices. One
common approach to interpret community structure represented by of monitoring data
in terms of species diversity. Either or both of two components may be included when
characterizing species diversity: species richness (i.e., the numbers of species within the
community) and species evenness (i.e., distribution of die relative abundances among
these species). As recently summarized by Crowder (1990), the diversity of species in
waterbodies can be determined by colonization rates, extinction rates, competition,
predation, physical disturbance, pollution, and other factors. A common problem in
assessments of monitoring data is to determine how these multiple factors interact to
determine species diversity at the monitored sites.
A useful two-step qualitative data assessment approach can help to unravel
determinants of species compositions in collected monitoring data. The first step again,
as discussed above for the qualitative examination of collected data, is to compare the
list of species collected and their relative population sizes against published records of
the known sensitivities by these taxa, or their close relatives, to various contaminants
7-48
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
known or suspected to be present The tendency of species to be abundant, present, or
absent relative to their individual relative tolerances or sensitivities to, for example,
sediments, temperature regimes, or various chemical pollutants should be noted.
Frequently, also as suggested above, this comparison will begin pointing toward the
most likely determinants of species diversity at the sampled sites.
TMDL developers should note two important cautions with respect such
comparisons. First, different strains of the same species sometimes can have widely
different sensitivities to a stressor. This is especially relevant for fish species that have
undergone extensive hatchery breeding programs. These efforts have affected
sensitivities not only for characteristics targeted by the breeding program, but also for
nontargeted sensitivities to other potential stressors. Second, when incorporating
information from indicator species lists into an assessment, it is important to review and
evaluate whether the rankings shown on the list include data collected from the region
of interest When the list includes no species from the area surrounding the receiving
water of concern, use of the listed species can represent an over-extension of the
credibility of the list. In any assessment using such lists, the TMDL developer should
document possible known or expected limitations of the lisKs) and note the possible
implications for the assessment.
The second step then involves integrating into this assessment implications of two
patterns often observed within distributions of species diversity values. Results from
two literature reviews indicate that two generalized relationships widely underlie
diversity patterns:
(1) Forces that disrupt ecosystems from an equilibrium or stable condition
(e.g., wind, irregular chronic toxic input) will tend to cause species
diversity to increase, unless the disturbance is too frequent, then diversity
will tend to decrease (Huston, 1979).
(2) Species diversities generally first increase to a maximum then decrease
again across many environmental gradients, including availabilities of
resource supplies such as nutrients (Marcus, 1980).
Knowledge of these two generalized patterns can sometimes help in interpreting
species diversity patterns defined through monitoring data for spedes numbers,
abundances, and sensitivities to environmental stressors. Potential causes for
relationships appearing in the species diversity data also can be suggested. Such
relationships can provide the basis for establishing testable hypotheses for developing
and implementing special studies targeted to better define source-pollutant-impact
relationships.
7-49
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
A final caution is necessary regarding some assessment approaches using species
diversity. A number of indices have been developed to characterize species diversity
and evenness. Many of these are presented in the various guidance documents cited
below. Computed indices often include various potentially intractable problems for
statistical analysis (see Green, 1979; Sokal and Rohlf, 1981). Frequent among these
problems for species diversity indices are extremely wide variance estimates and, for
some, an inability to calculate variances. Potential deficiencies in the indices used should
be noted and, if possible, assessed as part of the overall study and analysis design and
assessment effort.
While diversity indices can provide useful general indicators of environmental
effects, investigators should generally limit their use to within-study comparisons, where
sampling and sample analysis methods are consistent. Additionally, if an assessment
uses a diversity index, the TMDL developer should also demonstrate an understanding
and (when possible) provide an evaluation of potential limitations caused by using the
index. Hurlbert (1971), Peet (1974), and Pielou (1975,1977) provide additional discussion
on the use and misuse of species diversity indices.
7.6.3 Algae
EPA's UAA manuals present instructive introductory guidance for interpreting
results from monitoring samples for algal communities collected from streams and
rivers, estuaries, and lakes and reservoirs (U.S. EPA, 1983,1984a, b). Additional useful
guidance is available in two additional EPA documents:
• Environmental Requirements and Pollution Tolerance of Freshwater Diatoms
(Lowe, 1974)
• Guide to the Identification, Environmental Requirements, and Pollution Tolerance
of Blue-Green Algae (Cyanophyta) (VanLandingham, 1982)
7.6.4 Benthic Macroinvertebrates
Interpretation of biological monitoring data for larger bottom-living invertebrates
is greatly aided using information contained in EPA's three UAA manuals (U.S. EPA,
1983,1984a, b). Their RBP manual also provides additional useful information to help
interpret these data collected from flowing water environments (Plafkin et al., 1989).
Most recently, EPA's manual for Macroinvertebrate Field and Laboratory Methods for
Evalmting the Biological Integrity of Surface Waters (Klemm et al, 1990) provides additional
useful guidance, including discussions and presentations of:
7-50
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
• Analysis of qualitative and quantitative data
• Community metrics and pollution indicators
• Pollution tolerance of selected macroinvertebrates
• Hilsenhoff's family level pollution tolerance values for aquatic arthropods
7.6.5 Fish
Beyond the various EPA documents noted above that also provide information
helpful when interpreting monitoring data for fish samples, two reference works
published by the American Fisheries Society contain a wealth of useful information. The
first, Fisheries Techniques (Nielsen and Johnson, 1983), focuses mainly on considerations
important for work in the field. Its chapters include discussions on advantageous and
shortcomings of most of the currently practiced sampling techniques. The document
discusses many potential problems associated with most field data collection techniques.
It also reviews methods and important considerations for length, weight and associated
structural indices, age determinations, fish diet analysis, and angler characterizations.
The companion volume, Methods for Fish Biology (Schreck and Moyle, 1990),
focuses primarily on methods used to analyze and assess collected fish samples in the
laboratory and in the office. As stated in that book's Preface,.. pros and cons of
alternative procedures are treated, as are uses and misuses of the data generated by the
techniques." Among the chapters are ones on research method designs, fish growth,
stress and acclimation, reproduction, behavior, population ecology, and community
ecology.
7.7 Physical Aquatic Habitat Evaluation
Section 101(a) of the Clean Water Act recognizes the importance of
preserving the physical integrity of the Nation's waterbodies, as discussed by U.S. EPA
(1983a). Physical habitat affects the types and numbers of species inhabiting a
waterbody. Characterizing physical habitats can help identify factors not related to
water quality that may impair die propagation and protection of aquatic life. Further,
this can help determine what uses can be attained in the waterbody given such
limitations. In general, physical properties such as temperature, water depth, currents
and flow, substrate, suspended solids, and reaeration rates can present critical limitations
that preclude attainment of beneficial uses. Hie physical characteristics of a waterbody
also greatly influence its reaction to pollutants and its natural purification processes.
7-51
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
Table 7-4 lists some of the key physical characteristics affecting habitats in the three
major categories of aquatic ecosystems.
Point and nonpoint sources can adversely affect physical habitats for aquatic
organisms through discharge of high-energy water flows and through discharge of
excess fine sediments. High-energy flows can erode bottom and bank materials,
destabilize the physical structure of aquatic habitats/ kill aquatic resident organisms, and
destroy eggs incubating in the benthic environment. Excess deposition of fine sediment
can coat bottom materials, smothering organisms and eggs that are still incubating.
EPA's UAA guidance for the various water types (U.S. EPA, 1983, 1984a, b) present
important considerations to aid in interpreting measurements of these and other physical
habitat variables.
Another source of potentially useful information to help interpret impacts to fish
associated with differences in physical habitat conditions are the Habitat Suitability
Index (HSI) Models developed by the U.S. Fish and Wildlife Service. Available HSI
models and sources to obtain copies were listed in Table 6-8.
7-52
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
Table 7-4.
Key physical characteristics affecting aquatic habitats
Ecosystem Type(s)
¦' ¦ ¦ i--- -iJM———====SSS
Important Characteristics
Reference(s)
Streams, rivers
Flow
Suspended solids
Sediments
Pool-riffle ratios
Run-bend ratios
Substrate composition
Channel characteristics
(including channelization)
Temperature
Bank stability
Riparian cover
U.S. EPA, 1983,
Plafkin et ai,
1989
Lakes,
impoundments
Size
Shape
Depth
Temperature
Flow/current regimes
U.S. EPA, 1984b
Estuaries
Tides
Wind shear
Momentum and buoyancy
of
freshwater inflows
Topographic frictional
resistance
Coriolis effect
Vertical mixing
Horizontal mixing
U.S. EPA, 1984a
7-53
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
References. Chapter 7
Baker, J.P., D.P. Bernard, M.J. Sale, and S.W. Christensen. 1990. Biological Effects of
Changes in Surface Water Add-base Chemistry. NAPAP Report 13. National
Acid Precipitation Assessment Program, Superintendent of Documents,
Government Printing Office, Washington, DC.
Barraclough, W.E., and D. Robertson. 1972. The fertilization of Great Central Lake. DOE.
Effect on juvenile sockeye salmon. Fisheries Bulletin 70:37-48.
Berg, L. 1982. The effect of exposure to short-term pulses of suspended sediment on
the behavior of juvenile salmonids. Pages 177-196 in G.F. Hartman, et al., editors.
Proceedings of the Carnation Creek workshop: A ten-year review. Department
of Fisheries and Oceans, Pacific Biological Station, Nanaimo, British Columbia,
Canada.
Berg, L., and T.G. Northcote. 1985.- Changes in territorial, gill-flaring, and feeding
behavior in juvenile coho salmon (Oncorhynckus kisutch) following short-term
pulses of suspended sediment Canadian Journal of Fisheries and Aquatic
Sciences 42:1410-1417.
Berryman, D., B. Bobee, D. Cluis, and J. Haemmerli. 1988. Nonparametric tests for
trend detection in water quality time series. Water Resources Bulletin 24:545-556.
Bisson, P.A., and R.E. Bilby. 1982. Avoidance of suspended sediment by juvenile coho
salmon. North American Journal of Fisheries Management 2:371-374.
Bjornn, T.C., M.A. Brusven, MP. Molnau, J.H. Milligan, R.A. Klamt, E. Chacho, and C.
Schaye. 1977a. Transport of granitic sediment in streams and its effects on
insects and fish. Bulletin No. 17. Forest, Range and Wildlife Experiment Station,
University of Idaho, Moscow, ID.
Black, J.A., R.F. Roberts, D.M. Johnson, D.D. Minicucci, K.H. Marcy, and M.E. Allen.
1975. The significance of physicochemical variables in aquatic bioassays of heavy
metals. Pages 259-275 in G.E. Glass, editor. Bioassay techniques and
environmental chemistry. Ann Arbor Science Publishers Inc. Ann Arbor, MI.
Box, G.E.P., W.G. Hunter and J.S. Hunter. 1978. Statistics for Experimenters: An
Introduction to Design, Data Analysis/ and Model Building. Wiley, New York.
7-54
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
Bums, D.C., and R.E. Edwards. 1985. Embeddedness of salmonid habitat of selected
streams on the Payette National Forest. U.S. Department of Agriculture, Forest
Service, Payette National Forest, McCall, ID.
Chapman, D.W. 1988. Critical review of variables used to define effects of fine
sediment in redds of large salmonids. Transactions of the American Fisheries
Society 117:1-21.
Chapman, D.W., and K.P. McLeod. 1987. Development of Criteria for Fine Sediment
in the Northern Rockies Ecoregion. EPA 910/9-87-162. Region 10, U.S.
Environmental Protection Agency, Seattle, WA.
Cox, D.C.and P.C. Baybutt 1981. Methods for uncertainty analysis: A comparative
survey. Risk Analysis 1(4): 251-258.
Crowder, L.B. 1990. Community Ecology. Pages 609-632 in C.B. Schreck and P.B.
Moyle (editors). 1990. Methods for Fish Biology. American Fisheries Society,
Bethesda, MD.
Davies, P.H., and J.D. Woodling. 1980. Importance of laboratory-derived metal toxicity
results in predicting in-stream response of resident salmonids. Pages 281-299 in
J.G. Eaton, P.R. Parrish, and A.C. Hendricks, editors. Aquatic Toxicology. ASTM
STP 707. American Society for Testing and Materials.
DeGroot, M.H. 1986. Probability and Statistics (2nd Edition). Addison-Wesley, Reading,
MA
Donigian, A.S. and W.C. Huber. 1991. Modeling of Nonpoint Source Water Quality in
Urban and Non-urban Areas. EPA/600/3-91/039. ERL Athens.
Everest, F.H., R.L. Beschta, C.J. Scrivener, et al. 1987. Fine sediment and salmonid
production: A paradox. Page 98-142 in E.O. Salo, and T.W. Cundy, editors.
Streamside management: Forestry and fishery interactions. Contribution 57.
College of Forest Resources, University of Washington, Seattle, WA.
Flavelle, P. 1992. A quantitative measure of model validation and its potential use for
regulatory purposes. Advances in Water Resources 15:5-13.
Fleet, R.J, D.W. Schindler, R.D. Hamilton, and N.E.R. Campbell. 1980. Nitrogen fixation
in Canadian Precambrian Shield Lakes. Canadian Journal of Fisheries and
Aquatic Sciences 37:494-505.
7-55
-------
The Cadmus Group> Inc.
DRAFT
June 10,1993
Freeze. R.A., J. Massmann, L. Smith, T. Sperling and B. James. 1990. Hydrogeological
Decision Analysis: 1. A Framework. Ground Water 28(5): 738-766.
Gallant, A.L., T.R. Whittier, D.P. Larsen, J.M. Omernik, and R.M. Hughes. 1989.
Regionalization as a Tool for Managing Environmental Resources. EPA/600/3-
89/060. Environmental Research Laboratory, U.S. Environmental Protection
Agency, Corvallis, OR.
Gilbert, R.O. 1987. Statistical Methods for Environmental Pollution Monitoring. Van
Nostrand Reinhold, New York
Golterman, H.L. 1975. Physiological limnology, an approach to the physiology of lake
ecosystems. Elsevier Scientific Publishing Company, New York, NY. 489 pp.
Green, R.H. 1979. Sampling Design and Statistical Methods for Environmental
Biologists. John Wiley and Sons, New York.
Hasfurther, V.R. 1985. The use of meander parameters in restoring hydrologic balance
to reclaimed stream beds. Pages 21-40 in J.A. Gore, editor. The restoration of
rivers and streams: Theories and experience. Butterworth Publishers, Stoneham,
MA.
Heede, B.H. 1980. Stream dynamics: An overview for land managers. General
Technical Report RM-72. Rocky Mountain Forest and Range Experiment Station,
U.S. Department of Agriculture, Forest Service, Fort Collins, CO. 26 p.
Herbert, D.W.M., and J.C. Merkens. 1961. The effect of suspended mineral solids on the
survival of trout. International Journal of Air and Water Pollution 5:46-55.
Hill, A.R., 1986. Stream nitrate-N loads in relation to variations in annual and
seasonal runoff regimes. Water Resources Bull. 22: 829-839.
Hipel, K.H. (editor). 1988. Nonparametric Approaches to Environmental Impact,
(reprinted from Water Resources Bulletin, Volume 24, Number 3). American
Water Resources Association, Bethesda, MD.
Hollander, M. and D.A. Wolfe. 1973. Nonparametric Statistical Methods. Wiley, New
York.
Horner, R.R., E.B. Welch and R.B. Veenstra. 1983. Development of nuisance periphytic
algae in laboratory streams in relation to enrichment and velocity, pp. 121-124 in
Wetzel, R.G. (ed.), Periphyton of Freshwater Ecosystems, Dr W. Junk Publishers,
7-56
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
The Hague.
Hurlbert, S.H. 1971. The nonconcept of species diversity: a critique and alternative
parameters. Ecology 52:577-586.
Huston, M. 1979. A general hypothesis of species diversity. American Naturalist
113:81-101.
Hutchinson, G.E. 1969. Eutrophication, past and present. Pages 17-26 in
Eutrophication: Causes, Consequences, Correctives, Proceedings of a symposium.
National Academy of Sciences, Washington, District of Columbia.
Hutchinson, G.E. 1973. Eutrophication - The scientific background of a contemporary
practical problem. American Scientist 61:269-279.
Hynes, H.BJN. 1969. The enrichment of streams. Pages 188-196 in Eutrophication:
Causes, Consequences, Correctives, Proceedings of a symposium. National
Academy of Sciences, Washington, District of Columbia.
Hynes, H.B.N. 1970. The ecology of running waters. University of Toronto Press,
Ontario, Canada. 555 pp.
Inman, R.L. and J.C. Helton. 1988. An investigation of uncertainty and sensitivity
analysis techniques for computer models. Risk Analysis 8(1): 71-90.
International Atomic Energy Agency (IAEA). 1989. Evaluating the reliability of
predictions made using environmental transfer models. Environmental Safety
Series No. 100. IAEA, Vienna.
Jewell, T.K., T.J. Nunno and D.D. Adrian. 1978. Methodology for calibrating storoiwater
models. Journal of the Environmental Engineering Division, ASCE, 104:485.
Kendall, M.G. 1975. Rank Correlation Methods (4th ed.). Charles Griffin, London.
Klemm, D.J., P.A. Lewis, F. Fulk, and J.M. Lazorchak. 1990. Macroinvertebrate Field
and Laboratory Methods for Evaluating the Biological Integrity of Surface Waters.
EPA/600/4-90-030. Environmental Monitoring Systems Laboratory, U.S.
Environmental Protection Agency, Cincinnati, OH.
Likens, G.E., F. H. Borman, R.S. Pierce, J.S. Eaton, and N.M. Johnson, 1977.
Biogeochemistry of a forested ecosystem. Springer-Verlag, New York.
7-57
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
Lloyd, D.S. 1987. Turbidity as a water quality standard for salmonid habitats in Alaska.
North American Journal of Fisheries Management 7:43-45.
Lloyd, D.S., J.P. Koenings, and J.D. LaPerriere. 1987. Effects of turbidity in fresh waters
of Alaska. North American Journal of Fisheries Management 7:18-33.
Loftis, J.C., R.C. Ward, R.D. Phillips, and C.H. Taylor. 1989. An Evaluation of Trend
Detection Techniques for Use in Water Quality Monitoring Programs.
EPA/600/3-89/037. Office of Acid Deposition, Environmental Monitoring, and
Quality Assurance, U.S. Environmental Protection Agency, Washington, DC.
Loftis, J.C., C.H. Taylor, and P.L. Chapman. 1991. Multivariate Tests for Trend in Water
Quality. Water Resources Research 27:1419-1429.
Lohman, K., J.R. Jones, and B.D. Perkins. 1992. Effects of nutrient enrichment and flood
frequency on pexiphyton biomass in northern Ozark streams. Canadian Journal
of Fisheries and Aquatic Science 49:1198-1205.
Lowe, R.L. 1974. Environmental Requirements and Pollution Tolerance of Freshwater
Diatoms. EPA-670/4-74-005. National Environmental Research Center, U.S.
Environmental Protection Agency, Cincinnati, OH.
Mackenthun, K.M. 1969. The Practice of Water Pollution Biology. Division of Technical
Support, Federal Water Pollution Control Administration, Department of Interior,
Washington, DC.
Marcus, M.D. 1980. Periphytic community response to chronic nutrient enrichment by
a reservoir discharge. Ecology 61:387-399.
Marcus, M.D., B.R. Parkhurst, J.P. Baker, C.S. Creager, T.E. Fannin, C.G. Ingersoll, D.R.
Mount, and F.J. Rahel. 1986. An evaluation and compilation of the reported
effects of acidification on aquatic biota. Volume 1: Compiled data. EA-4825,
Volume 1, Research Project 2346-1, Electric Power Research Institute, Palo Alto,
CA.
Marcus, M.D., M.K. Young, L.E. Noel, and B.A. Mullan.. 1990. Salmonid-Habitat
Relationships in the Western United States: A Review and Indexed Bibliography.
General Technical Report RM-188. Rocky Mountain Forest and Range Experiment
Station, U.S. Department of Agriculture, Forest Service, Fort Collins, CO.
Marin, C.M., M.A. Medina, and J.B. Butcher, J.B. 1989. Monte Carlo analysis and
Bayesian decision theory for assessing the effects of waste sites on ground water,
7-S8
-------
The Cadmus Group> Inc.
DRAFT
June 10,1993
I: theory. J. Contam. Hydrol. 5:1-13.
Martin, J.L., R.B. Ambrose and S.C. McCutcheon, eds. 1990. Technical Guidance Manual
for Performing Waste Load Allocations, Book ID: Estuaries, Part 2: Application of
Estuarine Waste Load Allocation Models. EPA-823-R-92-003. Office of Water
McCutcheon, S.C., Z. Dongwei and S. Bird. 1990. Model calibration, validation, and use.
Chapter 5 in J.L. Martin et al., eds., Technical Guidance Manual for Performing
Waste Load Allocations, Book HI: Estuaries, Part 2: Application of Estuarine Waste
Load Allocation Models. EPA-823-R-92-003. Office of Water
McLeod, A.I., K.W. Hipel, and F. Comancho. 1983. Trend assessment of water quality
time series. Water Resources Bulletin 19:573-547.
Montgomery, P.H., and K.H. Reckhow. 1984. Techniques for detecting trends in lake
water quality. Water Resources Bulletin 20:43-52.
Mount, D.R., C.G. Ingersoll, D.D. Gulley, J.D. Fernanadez, T.W. LaPoint, and H.L.
Bergman. 1988. Effect of long-term exposure to acid, aluminum, and low calcium
on adult brook trout (Salvelinus fontinalis). 1. Survival, growth, fecundity, and
progeny survival. Canadian Journal of Fisheries and Aquatic Sciences
45:1623-1632.
National Academy of Sciences (NAS). 1969. Eutrophication: Causes, Consequences,
Correctives, Proceedings of a symposium. National Academy of Sciences,
Washington, District of Columbia.
Nielsen, L.A., and D.L. Johnson (editors). 1983. Fisheries Techniques. American
Fisheries Society, Bethesda, MD.
Nix, S.J., P.E. Moffa and DP. Davis. 1991. The practice of combined sewer system
modeling. Water Resources Bulletin 27(2): 189-200.
Ohio EPA. 1988. Biological Criteria for the Protection of Aquatic Life: Volume II: Users
Manual for Biological Held Assessment of Ohio Surface Waters. Ohio
Environmental Protection Agency, Division of Water Quality Planning and
Assessment, Columbus, OH.
Olson, W.H., D.L. Chase, and JJSF. Hanson. 1973. Preliminary studies using synthetic
polymer to reduce turbidity in a hatchery water supply. Progressive Fish-
Culturist 35:66-73.
7-59
-------
The Cadmus Group, Inc.
DRAFT
June 10,1993
Pankratz, A. 1983. Forecasting with Univariate Box-Jenkins Models. John Wiley and
Sons, New York, NY.
Parkhurst, B.R., R.G. Elder, J.S. Meyer, D.A. Sanchez, R.W. Pennak, and W.T. Waller.
1984. An environmental hazard evaluation of uranium in a Rocky Mountain
stream. Environmental Toxicology and Chemistry 3:113-124.
Parrish, R.S. and C.N. Smith. 1990. A method for testing whether model predictions fall
within a prescribed factor of true values, with an application to pesticide leaching.
Ecological Modeling, 51: 59-72.
Pascoe, D., S.A. Evans, and J. Woodworth. 1986. Heavy metal toxicity to fish and the
influence of water hardness. Archives of Environmental Contamination and
Toxicology 15:481-487.
Peet, R.K. 1974. The Measurement of Species Diversity. Annual Review of Ecology and
Systematics. 5:285-307.
Pielou, E.C. 1975. Ecological Diversity. Wiley-Intersdence, New York, NY.
Pielou, E.C. 1977. Mathematical Ecology. Wiley-Intersdence, New York, NY.
Plafkin, J.L., M.T. Barbour, K.D. Porter, S.K. Gross, and R.M. Hughes. 1989. Rapid
Bioassessment Protocols for Use in Streams and Rivers. Benthic
Macroinvertebrates and Fish. EPA/440/4-89/001. Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
Reckhow, K.H. and S.C. Chapra. 1983. Confirmation of water quality models. Ecological
Modelling, 20:113-133.
Reckhow, K.H., J.T. Clements and R.C. Dodd. 1990. Statistical evaluation of mechanistic
water-quality models. J. Environ. Eng. 116(2): 250-268.
Schofield, C.L., and J.R. Trojnar. 1980. Aluminum toxicity to brook trout (Salvelinus
fontinalis) in acidified waters. Pages 341-362 in T.Y. Toribara, M.W. Miller, and
P.E. Morrow, editors. Polluted Rain. Plenum Press,.NY.
Schreck, C.B. and P.B. Moyle (editors). 1990. Methods for Fish Biology. American
Fisheries Society, Bethesda, MD.
Shoemaker, L.L. et al. 1992. Compendium of Watershed-Scale Models for TMDL
Development EPA/841-R-92-002. U.S. EPA Office of Wetlands, Oceans and
7-60
-------
The Cadmus Group; Inc.
DRAFT
June 10,1993
Watersheds and Office of Science and Technology.
Sidle, R.C. 1988. Bed load transport regime of a small forest stream. Water Resources
Research 24(2):207-218.
Sidle, R.C., and A.J. Campbell. 1985. Patterns of suspended sediment transport in a
coastal Alaska stream. Water Resources Bulletin 21(6):909-917.
Sigler, J.W., T.C. Bjornn, and F.H. Everest. 1984. Effects of chronic turbidity on density
and growth of steelheads and coho salmon. Transactions of the American
Fisheries Society 113:142-150.
Smith, V.H. 1983. Low nitrogen to phosphorus ratios favor dominance by blue-green
algae in lake phytoplankton. Science 221:669-671.
Snedecor, G.W. and W.G. Cochran. 1980. Statistical Methods (7th ed.). Iowa State
University Press, Ames, IA.
Sokal, R.R. and F.J, Rohlf. 1981. Biometry, the Principles and Practice of Statistics in
Biological Research. W.H. Freeman and Company, San Francisco, CA.
Spooner, J., R.P. Maas, S.A. Dressing, M.D. Smolen, and F.J. Humenik. 1985. Appropriate
Designs for Documenting Water Quality Improvements from Agricultural
Nonpoint Source Programs. In Perspectives on Nonpoint Pollution. U.S.
Environmental Protection Agency, Washington, DC. EPA 440/5-85-001.
Stockner, J.G. 1981. Whole-lake fertilization for the enhancement for sockeye salmon
(Oncorhynchus nerka) in British Columbia, Canada. Verh. Internat. Verin. Limnol.
21:293-299.
Striffler, W.D. 1965. The Selection of Experimental Watersheds and Methods in
Disturbed Forest Areas. In Publication No. 66 of the I.A.SJ1. Symposium of Budapest.
Stumm, W., and J.J. Morgan. 1970. Aquatic chemistry, an introduction emphasizing
chemical equilibria in natural waters. Wiley-Intersdence, A Division of John
Wiley and Sons, Inc. New York, NY. 583 pp.
Sykora, J.L., E.J. Smith, and M. Synak. 1972. Effect of lime neutralized iron hydroxide
suspensions on juvenile brook trout (Salvelinus fontinalis Mitchell). Water
Research 6:935-950.
Taylor, C.H., J.C. Loftis. 1989. Testing for trend in lake and ground water quality time
7-61
-------
The Cadmus Group> Inc.
DRAFT
June 10,1993
series. Water Resources Bulletin 25:715-726.
Thomann, R.V. 1980. Measures of verification, in Workshop on Verification of Water
Quality Models. EPA-600/9-80-016. ERL Athens, GA
Thurston, R.V., R.C. Russo, C.M. Fetterolf, T.A. Edsall, and Y.M. Barber Jr., editor. 1979.
Review of the EPA Red Book: Quality criteria for water. Water Quality Section,
American Fisheries Society, Bethesda, MD. 313 pp.
Tuden, R., D. Smith, and M. Rea. June 17,1992. Guidance for modifying water quality
standards and protecting effluent-dependent ecosystems. Interim Final. EPA
Region 9, San Francisco, CA.
U.S. Environmental Protection Agency. 1983. Technical Support Manual: Waterbody
Surveys and Assessments for Conducting Use Attainability Analyses. Office of
Water, U.S. Environmental Protection Agency, Washington, DC.
U.S. Environmental Protection Agency. 1984a. Technical Support Manual: Waterbody
Surveys and Assessments for Conducting Use Attainability Analyses, Volume II:
Estuarine Systems. Office of Water, U.S. Environmental Protection Agency,
Washington, DC.
U.S. Environmental Protection Agency. 1984b. Technical Support Manual: Waterbody
Surveys and Assessments for Conducting Use Attainability Analyses, Volume III:
Lake Systems. Office of Water, U.S. Environmental Protection Agency,
Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1986. Quality Criteria for Water
1986. EPA 440/5-86-001. Office of Water, U.S. Environmental Protection Agency,
Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1990. Biological Criteria, National
Program Guidance for Surface Waters. EPA-440/5-90-004. Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1991a. Guidance for Water Quality-
based Decisions: The TMDL Process. EPA 440/4-91-001. Assessment and
Watershed Protection Division.
U.S. Environmental Protection Agency (U.S. EPA). 1991b. Technical Support Document
for Water Quality-based Toxics Control. EPA/505/2-90-001. OWEP/OWRS.
7-62
-------
The Cadmus Group> Inc.
DRAFT
June 10,1993
U.S. Environmental Protection Agency (U.S. EPA). 1991c. Watershed Monitoring and
Reporting for Section 319 National Monitoring Program Projects. U.S.
Environmental Protection Agency, Office of Water, Washington, DC.
U.S. Environmental Protection Agency (U.S. EPA). 1993. Guidance Specifying
Management Measures for Sources of Nonpoint Pollution in Coastal Waters.
EPA-840-B-92-002. U.S. Environmental Protection Agency, Office of Water,
Washington, DC.
U.S. Environmental Protection Agency, Science Advisory Board (U.S. EPA SAB). 1991.
Evaluation of the Ecoregion Concept, Report of the Ecoregions Subcommittee of
the Ecological Processes and Effects Committee. EPA-SAB-EPEC-91-003. U.S.
Environmental Protection Agency, Washington, DC.
VanLandingham, S.L. 1982. Guide to the Identification, Environmental Requirements
and Pollution Tolerance of Freshwater Blue-Green Algae (Cyanophyta). EPA-
600/3-82-073. Environmental Monitoring and Support Laboratory, U.S.
Environmental Protection Agency, Cincinnati, OH.
Weber, C.I. (editor). 1973. Biological Field and Laboratory Methods for Measuring the
Quality of Surface Waters and Effluents. EPA-670/4-73-001. National
Environmental Research Center, U.S. Environmental Protection Agency,
Cincinnati, OH.
Wedepohl, R.E., D.R. Knauer, G.B. Wolbert, H. Olem, P.J. Garrison, and K. Kepford.
1990. Monitoring Lake and Reservoir Restoration. EPA 440/4-90-007. Prepared
by the North American Lake Management Society for Office of Water, U.S.
Environmental Protection Agency, Washington, DC.
Welch, E.B., R.R. Horner and C.R. Patmont. 1989. Prediction of nuisance periphytic
biomass: a management approach. Water Research 23(4): 401-405.
Welch, E.B., J.M. Jacoby, R.R. Horner and M.R. Seeley. 1988. Nuisance biomass levels of
periphytic algae in streams. Hydrobiologia 157:161-168.
Wetzel, R.G. 1975. Limnology. W.B. Saunders Company, Philadelphia, PA.
WPCF. 1989. Combined Sewer Overflow Pollution Abatement Manual of Practice No.
FD-17. Water Pollution Control Federation, Alexandria, VA.
7-63
-------
The Cadmus Group. Inc.
DRAFT
June 10.1993
Chapter VIII. Establishing TMDLs/WLAs/LAs and the MOS
Purpose: This chapter discusses the policies, procedures, and nuances of
translating data and model results into formally designated TMDLs, WLAs, LAs and an
MOS. As was discussed in chapter 1, the statutory mandate for TMDLs is found in
CWA § 303(d). However, the procedures for establishing TMDLS are regulated under
40 CFR 130.7. The TMDL Process guidance document (U.S. EPA, 1991a) provides a
detailed explanation of the States and EPA responsibilities for development and
submittals, but spends less time on the details of translating data and model results into
TMDLs. This guidance focuses on the latter.
8.1 Which Comes First, the TMDL or WLAs/LAs?
Logic would have it that the TMDL for a waterbody would be determined first
(i.e., determine the "size of the pie") based upon the system's assimilative capacity, and
then the TMDL ("pie") would be divided ("sliced") into portions reserved for WLAs and
LAs, and possibly even an explicit MOS. In the real world, however, assimilative
capacity is imprecisely known and often changes over space and time. It can be
dependent on season, variation in meteorological conditions, the distribution of pollutant
sources and their inherent characteristics, and a myriad of other ecological factors. Thus,
in many cases, it may not be feasible to use data analysis or modeling to arrive at a
single allowable load for a given pollutant that adequately and fairly addresses WQSs
and loading allocations.
The most practical method may be to use the model to evaluate various
combinations of WLAs and LAs that represent existing and potentially future source
scenarios. After all, by definition, the TMDL is the sum of all WLAs and LAs that serve
to protect WQSs in a given waterbody. Thus, the TMDL developer can use die model
in an iterative manner/ varying combinations of WLAs and LAs and noting which
alternatives protect WQSs throughout the watershed.' Combinations of WLAs and LAs
that are shown through the modeling analyses to protect WQSs are viable candidates for
the TMDL.
In some cases, there may not be a model or the level of uncertainty associated
with model results may be high. In these cases, an initial TMDL strategy of WLAs and
LAs based on best professional judgment may have to be implemented. Follow up
monitoring of the waterbody can demonstrate whether the TMDL is sufficient or needs
revision. Additional predictive modeling may be needed to support revision and
prevent an endless "trial and error" loop. However, even in the case with sophisticated
modeling, performance monitoring may indicate the need to "go bade to die drawing
board" with regard to the TMDL management strategy.
8-1
-------
The Cadmus Group. Inc.
DRAFT
June 10.1993
8.2 Deciding When to Employ the Phased Approach
The concept of the phased approach was discussed in chapter 1. There it was
stated that lack of adequate information would be cause to enter into a phased approach
that allowed for incremental building of the TMDL. However, most analysts would
agree that there is never perfect information on which to act. Thus, the issue arises for
the TMDL developer as to how to determine when the lack of information is substantial
enough to warrant a phased approach.
8.2.1 How Much is Enough?
Deciding when the lack of information is severe enough to employ a phased
approach to TMDL development is usually a judgment call There are, however, factors
that can be taken into consideration in making a sound decision. For one thing, the
TMDL developer can review any model reliability analyses that may have been
performed (see chapter 7 for more information on refiabilty analyses). If the reliability
analyses (formal or informal) reveal a high level of uncertainty in model results, then the
TMDL developer should scrutinize the impacts of potential error more carefully.
Sensitivity analyses for the model could demonstrate whether substantially different
results are predicted for key parameters where uncertainty is high. If these analyses
show that important decisions hinge on highly uncertain model assumptions, then the
TMDL developer may want to use a phased approach that allows for better information
to be gathered to reduce the uncertainty. Important decisions include those where the
viability of a designated waterbody use is threatened or where an action may be
required that exerts a high or disproportionate cost on one of the stakeholders.
8.2.2 Are Interim Control Measures Needed?
When a phased approach is chosen, the TMDL developer should decide whether
existing controls within the watershed are adequate for the intervening period.
"Adequate" in this case does not necessarily mean protection of WQSs, since the reason
that a TMDL strategy is being adopted will frequently be to restore an impaired
waterbody that is not being protected under current controls. Rather, it refers to a level
of protection that prevents severe threats to human health or aquatic life, or irreparable
damage from occurring in the interim.
In addition, interim measures may be taken to "maintain the status quo" while
final goals are established. This can be particularly important for watersheds facing the
stress from rapid population growth and land development. An example of this
occurred in the Tar-Pamlico Basin in North Carolina. North Carolina environmental
officials classified the basin as nutrient sensitive due to severe eutrophication effects.
However, not enough information was available to determine the level of nutrient
8-2
-------
The Cadmus Group. Inc.
DRAFT
June 10.1993
reduction that would be necessary to meet WQSs. Therefore, the State implemented an
interim plan with less stringent requirements that were expected to prevent an increase
in nutrient loading to the sensitive estuary area while a hydrodynamic water quality
model could be developed to help establish final reduction targets (NCDEHNR, 1989).
8.2.3 Scheduling Phased Approach Activities
Section 303(d) of the CWA mandates TMDLs be adopted for water quality limited
waterbodies, and 40 CFR 130.7 requires that States identify those priority waters for
which TMDLs will be adopted over the next two years. Thus, in delaying adoption of
a TMDL for a prioritized waterbody, the administering agency should specify the
schedule for when the final TMDL will be established. The TMDL Process document
(U.S. EPA, 1991) states that the scheduling should coordinate all various activities
(monitoring, modeling, permitting, etc.) and stakeholders (e.g., local, state, and federal
authorities, permittees, etc.).
Steps to Scheduling a Phased TMDL
Several steps may need to be gone
through in order to arrive at a reasonable
schedule. After determining what
additional information needs to be
gathered and the subsequent tasks that
should follow to fill those gaps and
complete the TMDL, the parties that are
involved (i.e., stakeholders) and their
roles/responsibilities should be identified.
This would include any agencies/
permittees/etc. that will be collecting
data, developing models, and/or
implementing follow up control measures.
Next, an estimate of resources needed to
perform the tasks should be developed
along with an inventory of available
resources. If there is a mismatch between available resources and those needed to
perform the tasks, then additional funding sources may need to be sought or the tasks
can be revised to work within resource constraints. Where possible, it may be
advantageous to spread the costs out over the stakeholders (i.e., pooling resources may
lead to a more efficient and effective information gathering effort). Finally, with tasks,
stakeholders, roles/responsibilities, and resources identified, a schedule of milestone
dates for completion of the tasks can be developed. Where sampling restrictions for
particular weather conditions exist, the schedule may need to be kept relatively flexible
to allow for time delays caused by inappropriate conditions.
•
Identify information gaps
•
Determine tasks needed to
•
complete TMDL
Identify stakeholder
~
roles/responsibilities
Estimate needed resources
v.UIIipart? llccUJj lU
resources
*
Obtain additional resources
or revise plan
•
Set milestone dates for
completion of tasks
8-3
-------
The Cadmus Grouv. Inc.
DRAFT
June 10.1993
8.3 Splitting Up the Pie
One of the most difficult tasks for the TMDL developer will undoubtedly be how
to divide up allocations for a given waterbody. While technical bases such as modeling
results play a key role in this task, policy considerations are inextricably intertwined.
Issues such as equitability, economics/ and "political will" often play a role, and decisions
will be made that are not always popular with all of the stakeholders. Nonetheless,
adoption of TMDLs requires that this task be performed. This section touches upon
some of the options and considerations facing TMDL coordinators in this regard.
8.3.1 Assigning a Margin of Safety (MOS)
Section 303(d) of the CWA requires TMDLs to include "a margin of safety which
takes into account any lack of knowledge concerning the relationship between effluent
limitations and water quality." Given that TMDLs address both point source (WLAs)
and nonpoint source (LAs) allocations, this concept may be extended to cover
uncertainty in BMP efficacy in addition to effluent limitations. With this in mind, TMDL
developers are faced with the issue of how to incorporate a MOS into the TMDL.
Two methods are discussed in the TMDL Process document (U.S. EPA, 1991).
The first and most commonly used method is to implicitly incorporate a margin of safety
through conservative model assumptions. Therefore, in cases where the TMDL
developer and modeler are not the same person, the TMDL developer should sit down
with the modeler and evaluate important model assumptions for their conservative
nature. This is not an easy task, however, for it requires a criterion for determining how
much of a safety factor is adequate. There is no "cut and dried" answer to that question.
Model uncertainty and sensitivity analyses may lend some insight; i-e., the greater the
uncertainty in sensitive parameters, the more margin of safety should be included in
estimating those parameters. If after scrutiny of the level of uncertainty in significant
model assumptions and the subsequent implications of error the TMDL developer
believes that the implicit MOS is not large enough, then either modeling assumptions
can be revised or an explicit MOS can be assigned.
An explicit MOS takes the form of a specified reserve of load. However, since the
TMDL is not likely to reflect a single numeric value, this reserve may have to be
associated with each component of a TMDL strategy. For instance, the reserve could
take a percentage off each WLA and LA. Instead of allocating 100 percent of the
allowable allocation, a portion could be withheld as a safety factor (e,g., component is
given 80 percent of the allowable allocation). In cases where a single TMDL value is
established, it would be simpler just to remove a portion of that load and use the
remaining portion to divide out into WLAs and LAs.
8-4
-------
The Cadmus Group. Inc.
DRAFT
Tune 10.1993
Establishing a MOS to account for the level of uncertainty can be problematic for
situations with a high level of uncertainty. There may be cases where the range of
possible error is large enough to cover an entire spectrum of control needs (i.ev from no
additional assimilative capacity available to the need for minimum control devices). In
these cases, being absolutely sure that the MOS will protect the waterbody would mean
holding the entire amount in reserve. This may be an option if the area is pristine and
designated an "outstanding resource water." However, in many cases there will already
be point sources and land use activities that require some reasonable level of allocation;
holding out the entire allocation to account for uncertainty would be unreasonable. In
these cases, a phased approach is recommended where additional information is sought
to reduce uncertainty. In addition, interim allocations can be assigned with a MOS that's
determined to be reasonable and a follow-up monitoring program can be implemented
to document effectiveness.
8.3.2 Achieving a Balance Between WLAs and LAs
During the early years of CWA implementation, EPA and States placed greatest
emphasis on the control of point sources through imposition of effluent limitations in
NPDES permits that were enforceable. This was a logical step since, in many cases,
municipal and industrial discharges existed without any form of treatment or with
minimal forms of treatment that did not perform at levels needed to protect WQSs.
Also, point sources are easier targets in that their effluent outfalls are readily identifiable
and can be monitored relatively easily. Many nonpoint sources, on the other hand, were
exempt from CWA permitting requirements and monitoring NFS is not typically an easy
task because of their diffuse nature.
With the 1987 amendments to the CWA, however, Congress began to recognize
the need to move beyond improving treatment technologies at point sources if the
nation's waters are expected to meet water quality objectives. Many States were
continuing to report impaired waters despite significant implementation of point source
controls. With the addition of section 319 of the Act, Congress set forth a mandate to
more aggressively implement best management practices (BMPs) that will reduce NPS
loads and aid in the protection of the nation's waters. In response to these actions by
Congress, EPA has placed a renewed emphasis on a watershed protection approach
(WPA) that attempts to establish a better balance between controls of point and nonpoint
sources.
The WPA makes the natural resource (i.e., the waterbody) the "client", rather than
a specific EPA or State agency or program. Thus, rather than simply trying to meet
specific minimal program or agency requirements, agencies (at all levels: federal, state,
and local) are being encouraged to take actions necessary to meet the "client's" needs.
Thus, assessments, prioritization, and management actions should strive to find the
8-5
-------
The Cadmus Grouv. Inc.
DRAFT
Tune 10.1993
appropriate balance of controls for both PS and NPS needed to meet WQSs within a
given waterbody.
Finding a balance between WLAs
and LAs in a TMDL management unit
involves evaluation of several factors.
First, the manager needs to know how
problem parameter loads are apportioned
among PS and NPS. Is one source
dominating the other? Imposition of
controls should reflect the size of the
source where possible. For instance, if a
pollutant load from NPS was found to be
80 percent of the total loading to a
problem area and a 40 percent overall
reduction in loading was needed, it
would make little sense to focus only on
point source controls.
Secondly, the TMDL developer
should look at the potential efficacy of
controls. What BMP and PS controls are
feasible and how effective will they be?
TMDL developers are unlikely to have
thorough knowledge in this regard and,
therefore, will likely need to seek input
from the stakeholders. Time constraints
may not allow for an indepth review in
every case, but efforts to gain an
understanding in the efficacy of feasible
controls will undoubtedly make for more
successful TMDL strategies. Helpful
guidance in specifying NPS management
U.S. EPA, 1993.
One argument that is often used
against balancing WLAs and LAs in
TMDL development is that LAs are
relatively unenforceable since they
involve some activities that are exempt
from permitting requirements and the
fact that BMP implementation is
overseen by nonregulatory agencies.
While this poses an implementation
challenge, it should not be made an
excuse for not assigning allocations that
require appropriate reductions from
sources of concern. Areas where
success cannot be achieved because of
the failure of NPS BMPs to be
implemented should be highlighted and
brought to the attention of those
involved. In some cases, new public
policy and/or regulations may be the
only way that LAs will be complied
with in the long run. However,
without attempts to include appropriate
LAs from the beginning, it is possible
that the problems will never be
addressed.
Enforcement of LAs
measures in coastal waters can be found in
Cost effectiveness should be considered. Since financial resources for controls are
limited, emphasis should be placed where possible on achieving the greatest return on
the money. For example, environmental officials in North Carolina recently evaluated
the needs for additional nutrient reduction controls in the Neuse River Basin due to
ongoing use impairment in the lower river and estuary portion of the basin. Point
Sources had already been addressed in the basin through imposition of phosphate bans
and effluent limitations such that their relative contribution of loading had been reduced
8-6
-------
The Cadmus Group. Inc.
DRAFT
Tune 10.1993
from approximately fifty percent down to around twenty percent in less than five years.
Rather than immediately considering additional PS controls, the State has chosen to
target specific watersheds for closer examination for improved implementation of BMPs.
Since reasonable LAs in the Neuse basin are currently unknown by the Water Quality
Section incharge of establishing the TMDLs for NC, the State has entered into a phased
TMDL approach to focus resources and schedules around the tasks of gathering enough
information to establish reliable LAs in addition to their existing WLAs.
As the level of information on targets, costs, removal efficiencies improves, formal
optimization analysis or multi-objective decision theory may be options for those familiar
with them. These methods would allow for decision criteria to be made explicit and,
therefore, might help those agencies looking for methods that appear less subjective to
stakeholders scrutinizing the allocation process.
8.3.3 Equitability and Fairness Between Allocations
One issue that arises in split-ting up the pie is that of equitability between
allocations. While there is no legal mandate that specifies that each allocation will be
of equal amount, its probably wise not to develop allocations that disproportionately
impact one stakeholder over another. This is not always an easy issue to decipher,
however, since there can be so many complicating factors such as differences between
sources in size, pollutant characteristics, design or layout, control technologies,
economics, and the list goes on....
Chadderton et al.(1981) provide some interesting examination of these issues
regarding methods to establish WLAs among interacting discharges. The following five
WLA methods were reviewed for a situation involving five interacting discharges of
BOD:
• Equal percent removal or equal percent treatment
• Equal effluent concentration
• Equal incremental cost above minimum treatment (normalized on the basis
of volumetric flow rate)
• Effluent concentration inversely proportional to pollutant mass inflow rate
• Modified optimization (i.e., least cost solution which includes the minimum
treatment requirements of the technology based controls)
A comparison of the methods was made based on cost, equity, efficient use of stream
assimilative capacity, and sensitivity to fundamental stream quality data. The authors
concluded that the "equal percent treatment" was preferable in the example stream
because of the method's insensitivity to data errors and accepted use by several states.
However, this is not to say that the other methods may not be preferable under different
8-7
-------
The Cadmus Grouv. Inc.
DRAFT
June 10.1993
circumstances or based upon other decision criteria.
8.3.4 Opportunities for Unique Solutions
Attempts to split up the pie often run into conflicts between stakeholders.
Everyone wants the largest slice they can have for fear of not having enough now or
sometime in the future. As environmental restrictions continue to be added, the
competition for allocations seems to increase. Thus, EPA and States are finding that the
time is ripe for unique solutions to the allocation dilemma.
Pollutant trading is one solution drawing considerable attention. The basic
premise of pollution trading is that it is more cost effective to reach the water quality
goals by targeting controls where they will do the most good. Therefore, rather than
applying across the board control requirements, participants are allowed to pool
resources and achieve targeted pollutant loading levels through the most cost effective
means possible.
Pollutant trading can occur between point sources (referred to as "point-point
trading") or between PS and NPS ("point-nonpoint trading"). Some examples of pilot
pollutant trading programs are provided in a document entitled, "Incentive Analysis for
Clean Water Act Reauthorization: Point Source/Nonpoint Source Trading for Nutrient
Discharge Reductions," (U.S. EPA, 1992a). In addition, a summary of the "Administator's
Point/Nonpoint Source Trading Initiative Meeting" was published by EPA (1992b).
While pollutant trading is not a tool that will work in every case, it represents a
concerted effort toward finding acceptable solutions to the issues arising from the
mandate to establish TMDLs, WLAs and LAs for water quality limited waterbodies.
8.3.5 Planning for the Future
The process of establishing TMDLs allows agencies a greater opportunity to plan
for the future. Via monitoring, data analysis and modeling, EPA and States are left with
better knowledge of the assimilative capacity of a system. Thus, when it comes time to
assign WLAs/LAs and a MOS, information regarding the current state of water quality
and activities within the watershed can be related to future considerations.
How much of the assimilative capacity is remaining? Regardless of the answer,
the information will be helpful to planning efforts of all stakeholders. For instance, if
assimilative capacity is exhausted in an area (a likelihood for impaired waters) and state-
of-the-art controls are already in place, then its known that additional growth in the
watershed will depend on either improving the state-of-the-art or source reduction.
While it may not be the answer some want to hear, it does give stakeholders valuable
8-8
-------
The Cadmus Grouv. Inc.
DRAFT
Tune 10.1993
information regarding where efforts for the future need to be targeted.
On the other hand, the TMDL process can be used to plan ahead to avoid early
exhaustion of the available assimilative capacity. Knowledge of future activities
regarding land-disturbing activities, population growth, etc. can be used to estimate
future wasteflows and NPS loads. Rather than allocating out all of the available
assimilative capacity to existing sources, it may be wiser to hold a reserve that covers
the future "build out" within a watershed. Given varying growth rates and land use
patterns, it may not always be reasonable or even feasible to carry out future analyses
to "build out" levels. But, planning for the future at some level may help to avoid future
conflicts caused by reallocations due to the failure to adequately consider future
activities. States using a basinwide planning approach have indicated that the basinwide
planning framework better allows for this future planning effort to take place (U.S. EPA,
1992c).
8.3.6 Defensibility of TMDL Allocations
Competition for allocations and die dislike of regulatory controls by some will
undoubtedly lead to the challenge of some TMDL allocation strategies. Therefore, in
order to stand up to such challenges, the TMDL developer should be sure that the
TMDL allocation strategy is defensible. Defensibility will be facilitated by three things:
sound science, good policy judgment, and good documentation. Given the legal
mandates to develop TMDLs, the burden of proof will be on those who challenge the
TMDL to show that it is inappropriate. Therefore, if the TMDL developer has used the
decision processes outlined in this document and based his/her recommendations in
sound science, and if the recommendations are supported in law or sound jpolicy, and
the process is well documented to demonstrate these facts, then the risk of overturning
a TMDL strategy should be reduced.
8.4 Translating WLAs into NPDES Permit Requirements
The NPDES permit is the mechanism for translating WLAs into enforceable
requirements for point sources. The National Pollutant Discharge Elimination System
(NPDES) is set forth in section 402 of the CWA. Under the NPDES program, permits
are required for the discharge of pollutants from most point source discharges into the
waters of the United States (see 40 CFR122 for applicability). While an NPDES permit
authorizes a point source facility to discharge, it also subjects the permittee to legally
enforceable requirements set forth in the permit.
8.4.1 Effluent Limitations
One of the ways in which WLAs are translated into permits is through the
8-9
-------
The Cadmus Grouv. Inc.
DRAFT
Tune 10. 1993
establishment of effluent limitations. Effluent limitations impose restrictions on the
quantities, discharge rates, and/or concentrations of specified pollutants in the point
source discharge. Effluent limitations reflect either minimum federal/state technology-
based guidelines or levels needed to protect water quality, whichever is more stringent.
By definition, TMDLs involve WLAs that are more stringent than technology-based
limits in order to protect WQSs and are therefore used to establish appropriate effluent
limitions.
As is required by 40 CFR 122.45(d), Converting WLAs into Permit Limits
effluent limitations are usually expressed
as some combination of daily maximum,
weekly average, and monthly average
concentrations. The limits chosen to fit
these formats often reflect separate
modeling analyses for different water
quality criteria. For example, daily
maximum limitations for toxics often
reflect near-field modeling analyses using
acute criteria, whereas longer term
average limits reflect well-mixed
conditions and chronic toxicity criteria.
Effluent limitations will not always be the same as the values for parameters
derived for WLAs. As stated in the Toxics TSD (U.S. EPA, 1991), "Direct use of a WLA
as a permit limit creates a significant risk that the WLA will be enforced incorrectly,
since effluent variability and the probability basis for the limit are not considered
specifically. For example, the use of a steady state WLA typically establishes a level of
effluent quality with the assumption that it is a value never to be exceeded. The same
values used directly as a permit limit could allow the WLA to be exceeded without
observing permit violations if compliance monitoring was infrequent." In this regard,
permit writers are encouraged to take a statistical approach to establishing limits from
WLAs.
In addition, WLAs often need to be converted to parameters that can be used
better for compliance judgment. For instance, modeling analyses often produce WLAs
for oxygen-demanding wastes in terms of ultimate CBOD and NBOD. However, these
parameters are difficult to use for compliance judgment because of the time and
complexity involved in measurement. Therefore, CBOD and NBOD model results are
converted via ratios to corresponding CBOD5 and NH3-N limitations which are easier
to measure and, because of their shorter time for analysis, allow for more rapid response
to any compliance problems that may arise.
• Consider regulatory format
requirements (e.g. daily max)
• Consider effluent variability and
probability basis for limit
• Convert WLA parameters to
compliance parameters where
necessary
8-10
-------
The Cadmus Grouo. Inc. DRAFT
8.4.2 Special Permit Conditions
Tune 10.1993
In addition to effluent limitations, some WLAs may need to be incorporated
through special conditions in the NPDES permit. This may be particularly true whore
non-numeric criteria are involved in the TMDL. Conditions can require specific modes
of operation or actions by the permittee to protect water quality, and these can be made
to address dynamic situations. For example, a discharger to a tidal area could be
required to discharge only during a specific portion of the tidal cycle to ensure complete
flushing or to reduce peak concentrations caused by pooling of the effluent.
Permit conditions can also be used to incorporate schedules for the permittee to
come into compliance with new limitations that are applied due to new or revised
TMDL/WLA requirements. Similarly, special conditions can be input to the permit to
require any special monitoring (including in the receiving waters) that is judged
necessary to ensure the adequacy of the TMDL/WLA and to document impacts on water
quality from the point source.
8.4.3 The Permit Issuance Process
NPDES permit requirements are not finalized until they have gone completely
through the permit issuance process. This includes periods for public review and, in the
case of major permits where states are the permit issuing authority, potential review by
EPA. Thus, it is possible that certain objections learned through the review process
could lead to the need for a change in permit requirements and, potentially, the TMDL
or WLA strategy.
8.5 Translating LAs into BMPs
Unlike NPDES permits for point sources, there are no corresponding pennit
requirements for nonpoint sources. Instead, load allocations are addressed, where
necessary, through implementation of best management practices (BMPs). In some cases,
states have certain mandatory BMP requirements for specific land use activities.
However, by in large, implementation of BMPs occurs through voluntary and incentive
programs such as government cost sharing.
While LAs may be used to target BMP implementation within a watershed,
translation of LAs into specific BMP implementation programs can be problematic. One
reason for this difficulty is that there are often many agencies involved in BMP
implementation. Rather than a single oversight agency, as is the case for NPDES
permits, BMP implementation can typically include federal, state, and local levels of
involvement. Many times the objectives of the varying agencies are different, which
makes coordination and enforcement difficult.
8-11
-------
The Cadmus Group, Inc.
DRAFT
Tune 10, 1993
In addition, as is reflected by much of the material in this document, it is not
always easy to accurately predict the effectiveness of BMPs. Therefore, its not easy to
determine how much of an effort to place into BMP implementation in order to comply
with LAs. As discussed earlier, TMDL strategies that are heavily dependent on loading
reductions through LAs are strong candidates for use of a phased approach because of
the inherent uncertainty in the system response to BMPs. A phased approach can
provide time for improved interagency coordination and establishment of long-term
watershed water quality monitoring programs to evaluate BMP effectiveness and
compliance with LAs.
8.6 Communicating the Results
In order for TMDL strategies to be successful, those parties likely to be effected
by the TMDL (i.e., the "stakeholders") should be involved in the TMDL development
process. Effective communication is a key element to the public participation process.
Stakeholders should be made aware of decisions regarding priority status of a
waterbody, modeling results or data analyses used to establish TMDLs for the
waterbody, and the pollutant control strategies resulting from the TMDL (i.e., WLAs and
LAs).
Methods for Communicating TMDLs
Public notices and follow up
meetings or hearings are one way in . , s
which stakeholders can be kept informed. r T ,j , ,, .
Federal regulations require public notice * "old ,P"b"c meetings/hearings
of NPDES permits and, therefore, those * C,ircul!te ba?? or watershed
notices can be used to convey additional plans for pubbc review
information about the TMDL process. * Use location and outreach
However, ifs best to try and involve the programs to expand general
public earlier on in the process in order to knowledge of TMDL process
avoid situations in which information is
learned by the permit issuing authority
that would have changed the way in which the TMDL process occurred. The TMDL
Process document (1992) recommends that the stakeholders be involved from the very
beginning, when the waters are being considered for listing on the 303(d) list along with
an assigned priority. Therefore, public notices of the 303(d) list are encouraged.
In addition, basinwide planning can facilitate communication and public
involvement. States that are organizing their program activities around a cyclical period
can schedule in public meetings throughout the process to keep interested stakeholders
informed. In addition, where a basin or watershed plan is prepared, the document can
be circulated for public review prior to adoption and implementation of the plan.
Typically, due to their comprehensive nature, basinwide plans also help communicate
8-12
-------
The Cadmus Group. Inc.
DRAFT
Tune 10.1993
TMDLs to the public because the plans provide a "big picture" that sets the context for
the TMDL. In this manner, the public learns what monitoring and assessment is being
performed, which parameters states or EPA are prioritizing for management, what
management alternatives have been considered, and which management strategy has
been selected for implementation.
Finally, where success of a TMDL strategy is dependent on other agencies for
implementation and enforcement of control mechanisms, it is critical that those agencies
be brought into the decision making process and soon as the need for their involvement
is realized. This means that communication typically must begin with these agencies
before final TMDL strategies are promulgated. The results of data and modeling
analyses that are used to begin considering management actions should be effectively
communicated to those agencies and the process of determining potential management
alternatives should then proceed jointly. Subsequently, TMDL results that are
communicated through a unified front of administering agencies more clearly
demonstrate to the other stakeholders that the requirements have been well planned and
coordinated.
References, Chapter VIII
NCDEHNR. 1992. Tar-Pamlico NSW Implementation Strategy. Division of
Environmental Management, Water Quality Section. Adopted December 14,1989, and
revised February 13,1992.
U.S. EPA. 1991a. Guidance for Water Quality-based Decisions: The TMDL Process. EPA
440/4-91-001. Assessment and Watershed Protection Division.
U.S. EPA. 1991b. Technical Support Document for Water Quality-based Toxics Control.
EPA/505/2-90-001. OWEP/OWRS.
U.S. EPA. 1992a. Incentive Analysis for Clean Water Act Reauthorization: Point
Source/Nonpoint Source Trading for Nutrient Discharge Reductions. Office of Water
and Office of Policy, Planning, and Evaluation. April 1992.
U.S. EPA. 1992b. Administrator's Point/Nonpoint Source Trading Initiative Meeting:
A Summary. EPA 841-S-92-001. Office of Water. August 1992.
U.S. EPA. 1992c. Watershed Cost Survey. Assessment and Watershed Protection
Division, Office of Water.
U.S. EPA. 1993. Guidance Specifying Management Measures for Sources of Nonpoint
Pollution in Coastal Waters. EPA 840-B-92-002. Office of Water, January 1993.
8-13
------- |