Assessing the TMDL Approach to Water Quality Management (2001)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

3 Waterbody Assessment: Listing and Delisting On July 27, 2000, the Assistant Administrator for Water at the U. S. Environmental Protection Agency (EPA) testified before a U.S. House committee that over 20,000 waterbodies across the United States were not meeting water quality standards according to Section 303d lists. Be- cause of legal, time, and resource pressures placed upon the states and EPA, there is considerable uncertainty about whether many of the waters on the 1998 303d lists are truly impaired. In many instances, waters pre- viously presented in a state’s 305b report5 or evaluated under the 319 Program6 were carried over to the state’s 303d list without any support- ing water quality data [e.g., see Iowa Senate File 2371, Sections 7–12 (Credible Data Legislation)]. Meanwhile, some waters that may be im- paired have yet to be identified and listed. The creation of an accurate and workable list of impaired waters is dependent on the first three steps of the Total Maximum Daily Load (TMDL) process, as depicted in Figure 1-1. States need to decide what waters should be assessed in the first place, how to create water quality standards for those waters, and then how to determine exceedance of 5 The Clean Water Act Section 305b report—the National Water Quality Inven- tory Report—is the primary vehicle for informing Congress and the public about general water quality conditions in the United States. This document character- izes water quality, identifies widespread water quality problems of national sig- nificance, and describes various programs implemented to restore and protect our waters (http://www.epa.gov/305b/). 6 Under the Clean Water Act Section 319 Nonpoint Source Management Pro- gram, States, Territories, and Indian Tribes receive grant money to support a wide variety of activities, including technical assistance, financial assistance, education, training, technology transfer, demonstration projects, and monitoring to assess the success of specific nonpoint source implementation projects (http://www.epa.gov/owow/nps/cwact.html). 32 

Waterbody Assessment: Listing and Delisting 33 those standards. Ideally, all these activities are encompassed and coordi- nated under the umbrella of a holistic ambient water quality monitoring program, described in the next section. However, given resource con- straints, the approaches currently used in most states to list impaired wa- ters fall short of this ideal. In recognition of these constraints, the com- mittee recommends changes to the TMDL program that would make the lists more accurate over the short and long terms. In addition, this chap- ter includes discussion on identifying waters to be assessed, defining measurable criteria for water quality standards, and interpreting moni- toring results for making the listing (and delisting) decision. ADEQUATE AMBIENT MONITORING AND ASSESSMENT The demands of an ambient-focused water quality management pro- gram, such as the TMDL program, require changing current approaches toward monitoring and assessment and subsequent decision-making. In many states, administrative performance measures (e.g., number of TMDLs developed, number of permits issued, and timeliness of actions) have been the principal measure of program effectiveness (Box 3-1). Such administrative measures are important, but reliance on such meas- ures diverts attention and resources away from environmental indicators of waterbody condition—the principal measures of effectiveness and success. Rather, information for decision-making should be based on carefully collected and interpreted monitoring data (Karr and Dudley, 1981; Yoder, 1997; Yoder and Rankin, 1998). The committee recognizes that state ambient monitoring programs have multiple objectives beyond the TMDL program (e.g., 305b reports, trends and loads assessments, and other legal requirements), which are not addressed in this report. It is suggested that to make efficient use of resources, states evaluate the extent to which their present ambient monitoring programs are coordi- nated and collectively satisfy their objectives. Ambient monitoring and assessment begins with the assignment of appropriate designated uses for waterbodies and measurable water qual- ity criteria that can be used to determine use attainment (EPA, 1995a). The criteria, which may include biological, chemical, and physical meas- ures, define the types of data to be collected and assessed. In response to the Government Performance and Results Act, the EPA Office of Water has developed national indicators for surface waters (EPA, 1995a) and a conceptual framework for using environmental information in decision- making (EPA, 1995b). EPA’s Office of Research and Development 

34 Assessing the TMDL Approach to Water Quality Management BOX 3-1 Ohio’s Experience with TMDLs In 1998, Ohio EPA’s Division of Surface Water (DSW) made recom- mendations for a process to develop TMDLs (Ohio EPA, 1999). The im- petus for developing a comprehensive TMDL strategy was (1) the na- tional attention brought about by lawsuits filed by environmental organi- zations and (2) the potential for the TMDL process to address all relevant sources of pollution to a waterbody. Prior to realizing the importance of this issue, state water quality management efforts were focusing on point sources and National Pollution Discharge Elimination System (NPDES) permitting, although since 1996, the leading cause of waterbody impair- ment has been shown to be nonpoint pollution and habitat degradation (Ohio EPA, 2000; Section 305b report). An agreement was reached between Ohio EPA and U.S. EPA Re- gion V on a 15-year schedule for TMDL development. Ohio’s 1998 303d list shows 881 of 5,000 waterbody segments as being impaired or threatened in 276 of the 326 watershed areas. Thus, completing TMDLs for all the currently listed segments by 2013 (in keeping with the 15-year schedule) will require an average of 18 watershed TMDLs per year as- suming that no new watersheds are added to future revisions of the 303d lists. It is understood that this latter assumption is unrealistic because a good portion of the state’s 5,000 waterbody segments has yet to be as- sessed, and it is a near certainty that additional waterbodies and water- sheds will be listed. Ohio recognizes that the technical and management processes required to implement TMDLs will need to go beyond the pur- view of the past emphasis on NPDES permits and point sources. At present, Ohio estimates it has sufficient resources available to de- velop only half of the TMDLs needed each year to produce the quality of product needed to meet various program expectations and expectations of stakeholders. Using 1998 as a baseline, approximately 16 percent of the DSW’s resources were dedicated to efforts that directly support TMDL development (see pie chart below). Without increases in funding, the resources will need to be diverted from other programs, or the pace of TMDL development will slow to the point where the 15-year schedule will need to be significantly extended. Diverting resources from other programs is highly unlikely in that each program faces unique chal- lenges, including reduction and elimination of NPDES permit backlogs continues 

Waterbody Assessment: Listing and Delisting 35 BOX 3-1 Continued and the growing need for new source permits, both of which place new burdens on the largest share of DSW resources. Devoting additional resources to TMDL development and implementation would require sig- nificant changes in water quality management emphasis on the national level, which seems unlikely given historical inertia and the emphasis placed on permitting programs by EPA and the states. Better coordina- tion between competing programs as well as additional resources are needed to resolve the present TMDL resource shortfall dilemma. Fo- cusing water quality management more on environmental results (as op- posed to administrative accomplishments alone) should provide a framework to better unify the emphasis and direction of competing pro- grams. recently published technical guidelines for the evaluation of ecological indicators (Jackson et al., 2000). One set of measurable parameters, termed indicators in Table 3-1, is offered for illustration. The core indi- cators include baseline biological, chemical, and physical parameters that comprise the basic attributes of aquatic ecosystems supplemented by specific chemical, physical, and bacteriological parameters from water, sediment, and tissue media, depending on the applicable designated use(s) and watershed-specific issues. Additional indicators not listed (e.g., biochemical markers and whole toxicity testing) may be appropri- ate as the situation dictates. 

TABLE 3-1. Core and Supplemental Indicators and Parameters that Comprise the Elements of an Adequate State Monitoring and Assessment Framework (after ITFM, 1992, and Yoder, 1997). Core Indicators Fish Macroinverte- Periphyton Physical habitat Chemical quality brates • Use at least two assemblages • Channel morphol- • pH ogy • Temperature • Flow regime • Conductivity • Substrate quality • DO • Riparian condition For Specific Designated Uses, add the following: Human/Wildlife Aquatic Life Recreation Water Supply Consumption Base list • Ionic strength • Fecal bacteria • Fecal bacteria • Metals (in tis- • Nutrients, sedi- • Ionic strength • Ionic strength sues) ment • Nutrients, • Organics sediment (in tissues) Supplemental • Metals • Other patho- • Metals list • Organics gens • Organics • Toxics • Organics • Other patho- gens 36 

More than one criterion may be necessary to determine attainment of a designated use, and each criterion will have strengths and limitations. In many instances of impairment—for example when riparian and aquatic habitats have been modified or flow regimes altered—biological parameters are better than chemical parameters at reflecting the condition of the aquatic ecosystem (Box 3-2). This is because biological assem- blages respond to and integrate all relevant chemical, physical, and bio- logical factors in the environment whether of natural or anthropogenic origin. On the other hand, relying only on biological assessments would not allow precise enough determination of associated causes and sources of impairments to satisfy water quality management needs including TMDL development. Over the long term, a full complement of meas- ured parameters must be the goal for water quality monitoring, assessing BOX 3-2 The Information Value of Monitoring Multiple Criteria The tendency for misdiagnosis of impairment by relying on only one type of criterion was illustrated in a study of more than 2,500 paired stream and river sampling sites in Ohio (Ohio EPA, 1990; Rankin and Yoder, 1990). In 51.6 percent of the samples, the results from biomoni- toring and chemical monitoring agreed—that is, they both detected either impairment or attainment of the water quality standard. This was par- ticularly true for certain classes of chemicals (e.g., toxicants), where an exceedance as measured by the chemical parameter was always asso- ciated with a biocriteria impairment. However, in 41.1 percent of the samples, impairment was revealed by exceedance of the biocriteria but not by exceedance of the chemical criteria. These results suggest that impairment may go unreported in areas where only chemical measure- ments are made. Interestingly, in 6.7 percent of the samples, chemical assessment revealed impairment that was not detected by bioassess- ment (especially for parameters such as ammonia-N, dissolved oxygen (DO), and occasionally copper). This latter occurrence is likely related to the fact that biocriteria have been stratified to reflect regional or ecotype peculiarities, and the more generically derived chemical criteria have not. Both the under- and overprotective tendencies of a chemical-criteria-only approach to water quality management can be ameliorated by joint use of chemical criteria and biocriteria, each used within their most appropri- ate indicator roles and within an adequate monitoring and assessment framework. 37 

38 Assessing the TMDL Approach to Water Quality Management chemistry and biology in a complementary manner and in their most ap- propriate indicator role (Karr, 1991; ITFM 1992, 1993, 1995; Yoder, 1997; Yoder and Rankin, 1998). At present, monitoring resources available to some states often do not allow for collecting and interpreting data for such a comprehensive suite of parameters. Indeed, ITFM (1995) reported that of the funding allocated by state and federal agencies to water quality management ac- tivities, only 0.2 percent was devoted to ambient monitoring. GAO (2000) has also noted the lack of adequate state budgets for the collection of meaningful data and for data interpretation. In response to these re- source shortfalls, the tendency has been to use only a single indicator of ambient conditions and often just a limited number of observations. Al- though some parameters can be monitored at lower costs than others, all monitoring can be costly (Yoder and Rankin, 1995). After standards development, a second requirement is adoption of a strategic and consistent approach to sampling and assessment given lim- ited data collection resources. Currently, the states use vastly different frameworks for monitoring and assessment, the net result of which is widely divergent estimates of the extent of impaired waters and of the proportion of waters that are fully assessed. This casts a great deal of uncertainty not only about what water quality problems are the most im- portant, but also about the accuracy and completeness of their delinea- tion. Errors in these estimates often become evident in the poor credibil- ity of 303d listings. A monitoring strategy that has promise in this limited-resource envi- ronment is the rotating basin approach, commonly referred to as a five- year basin approach (ITFM, 1995). As discussed in Box 3-3 for Florida, this approach is already followed by a number of states, at least in how ambient monitoring is accomplished7. As part of a rotating basin ap- proach, individual waters are assessed at differing levels of complexity each year, allowing for localized problems to be identified and solutions to be developed. For example, whether an individual assessment consists of an initial screening to identify gross impairment or a full assessment with more serious consequences will depend on how the information is to be used (for 305b reports, 303d listing, or other water quality pro- grams). Over time, different waterbodies are intensively studied as part of the rotation. Data collected can be used to support a number of differ- 7 In some states, the rotating basin approach is considered to be part of the ambi- ent monitoring program, while in others, it is a separate program. This report assumes the former throughout. 

Waterbody Assessment: Listing and Delisting 39 BOX 3-3 The Rotating Basin Program in Florida Settlement of a lawsuit brought by Earthjustice against EPA for its fail- ure to enforce timely actions to accomplish TMDL-related activities in Florida occurred in June 1999. Under the consent decree’s (CD) “Terms of the Agreement,” nearly 2,000 TMDLs in 711 waterbody segments are to be completed by the year 2011. Florida Department of Environmental Protec- tion (FDEP) has been named the lead agency to produce and adopt TMDLs, but its efforts must be coordinated with numerous other state and local agencies. In addition, the state has created opportunities for public participation throughout the TMDL generation and adoption process. To address the challenge of conducting the TMDL program and to better allocate its available resources, on July 1, 2000, Florida moved to the rotating basin approach for watershed management. Florida’s rotating basin approach has five phases (see below), with each phase taking about one year to complete. Further, FDEP has divided the state into 30 areas based on 8-digit hydrologic unit codes (HUCs), such that six areas repre- senting approximately one-fifth of Florida will be in the TMDL adoption phase in any one year. To meet the timelines ordered in the CD for Florida, FDEP must limit the time, effort, and resources it can commit in any one phase or waterbody. Because EPA has largely focused on addressing point source dis- charges through the NPDES permitting program, state and local govern- ments have in many cases taken the lead in dealing with nonpoint source issues, usually outside of the TMDL program. These programs often provide a flexible option to the time and budget constraints mentioned above. Florida believes that if local stakeholders are willing to initiate substantive programs that can fully, or even partially, accomplish the goals of the TMDL program at an expedited pace, then state and federal agen- cies should be able to support these actions, rather than delay or resist them. For example, in southwest Florida, a group of concerned stakeholders combined to form a “Nitrogen Consortium” (NC) to reduce inputs of nitrogen from all sources to the waters of Tampa Bay. Working together with the Tampa Bay Estuary Program and the FDEP, the NC developed a plan designed to “hold the line” against future increases of nitrogen (Tampa Bay National Estuary Program, 1996). Specific load- reduction efforts have been identified within the basin that allow for antici- pated growth to occur without resulting in a net increase in nitrogen loads to Tampa Bay. As would be anticipated under the conditions of a more formal TMDL, periodic reviews are made of the underlying assumptions and models used to further refine the nitrogen loads and associated goals. Although FDEP has not formally adopted a TMDL for Tampa Bay, EPA has approved these “hold the line” limits as a TMDL for Tampa Bay. continues 

40 Assessing the TMDL Approach to Water Quality Management BOX 3-3 Continued Florida’s Basin Management Cycle: 5 phases What happens in this phase? When does it occur? Build basin management team Prepare Status Report Phase I - Document physical setting Preliminary - Conduct water quality & TMDL assessments Years 1-2 Basin - Inventory existing & proposed Assessment management activities - Identify & prioritize management goals & objectives, & issues of concern - Develop Plan of Study Phase II Strategic Carry out strategic monitoring to collect Years 1-3 Monitoring additional data Compile & evaluate new data Finalize list of waters requiring TMDL Phase III Years 2-4 Develop TMDL Data Analysis Identify additional data collection needs & TMDL Report new findings Development Finalize management goals & objectives Develop draft Management Action Plan Identify monitoring & management partnerships, needed rule changes, Years 4-5 Phase IV legislative actions, and funding Management opportunities Action Plan Obtain participants’ commitment to implement plan Develop Monitoring & Evaluation Plan Implement Management Action Plan Secure project funding Carry out rule development/legislative action Year 5+ Phase V Transfer information to public & other Implementation agencies Conduct environmental education Monitor & evaluate implementation of plan 

Waterbody Assessment: Listing and Delisting 41 ent reporting and planning requirements, including a finding of attain- ment of water quality standards, a determination of impairment, or possi- ble delisting if the waterbody is found not to be impaired. Initial assess- ments that identify a waterbody as potentially impaired could be fol- lowed up by more thorough assessment. The rotating basin approach is an iterative process where the end result is both continual improvement of water quality management tools and policies and the ability to respond to emerging issues. Conclusions and Recommendations 1. To achieve the goal of ambient-based water quality manage- ment, monitoring and reporting must mature to focus on the condi- tion of the environment as the principal measure of success rather than on administrative measures. 2. Biological parameters should be used in conjunction with physical and chemical parameters to assess the condition of water- bodies. The use of both biological and chemical parameters is needed because they provide different and complementary types of information about the source and extent of impairment. 3. Evidence suggests that limited budgets are preventing the states from monitoring for a full suite of indicators to assess the con- dition of their waters and from embracing a rotating basin approach to water quality management. Currently, EPA is assessing the suffi- ciency of state resources to develop and implement TMDLs. Depending on the results of that assessment, Congress might consider aiding the states, for example through matching grants to improve data collection and analysis. EPA would be instructed to develop guidelines for such a program, if needed, making eligibility contingent on an approved state- wide monitoring and assessment strategy. 4. To allow states to better target limited monitoring budgets, EPA should set the TMDL calendar in concert with each state’s ro- tating basin program. The rotating basin approach used by several states is an excellent example of a rigorous approach to ambient moni- toring and data collection that can be used to conduct waterbody assess- ments of varying levels of complexity. For example, this approach can be used to create 305b reports, to list impaired waters, and to develop 

42 Assessing the TMDL Approach to Water Quality Management TMDLs. Once TMDLs are developed, the rotating basin approach could allow state and local governments to issue permits and implement man- agement programs based on the TMDLs in a coordinated manner. DEFINING ALL WATERS As shown in Figure 1-1, the TMDL process begins with identifica- tion of all waters for which achievement of water quality standards is to be assessed. The proposed regulations for the TMDL program (EPA, 1999a) define a waterbody as “a geographically defined portion of navi- gable waters, waters of the contiguous zone, and ocean waters under the jurisdiction of the United States, including segments of rivers, streams, lakes, wetlands, coastal waters and ocean waters.” The proposed regula- tions also require that states identify the geographic location of listed waterbodies using a “nationally recognized georeferencing system as agreed to by [the state] and the EPA.” States identify listed waterbodies using a variety of georeferencing systems, including stream segments in the EPA’s reach file system and watersheds in the U. S. Geological Sur- vey (USGS) system of hydrologic drainage basins. The use of such sys- tems for documenting the location of listed waters is convenient and pro- vides a degree of national standardization to the TMDL process. How- ever, the selection of a georeferencing system and a spatial scale for de- fining the totality of state waters is a more complicated issue (aside from the policy issue of national standardization). The EPA’s definition of waterbody implies that all state waters should be considered in the search for impaired waters and provides no guidance on a practical upstream limit or spatial scale to observe in that search. In theory, the hierarchy of tributaries in a watershed extends up- stream indefinitely. In practice, however, the choice of a lower limit on spatial scale or stream size has a very large influence on the total number of stream miles and small lakes that are included in the definition of state waters and thus require some form of assessment. For example, RF1, the original version of the EPA’s national reach file system (DeWald et al., 1985) contained approximately 65,000 stream reaches totaling approxi- mately 1 million km of stream channels. Now considered by EPA to be inadequate for describing the nation’s river and stream system, RF1 has been replaced by the National Hydrography Dataset (NHD) containing more than 3 million reaches totaling nearly 10 million km of channels. Moreover, a number of states have petitioned the EPA to add still lower- 

Waterbody Assessment: Listing and Delisting 43 order reaches (i.e., smaller streams) to the NHD in order to document the location of waters assessed by local interest groups. Because of local pressure and the lack of a regulatory lower limit on the size of streams and lakes to be considered, and because Geographic Information Systems (GIS) can document the existence and location of very small streams and lakes, the task of accurately and comprehensively assessing state waters has become formidable. At the current NHD scale, states contain an av- erage of about 70,000 stream reaches (>100,000 km), and given recent trends, that average is rising. This raises the question of how large the region of validity (the spa- tial area over which the data apply) is for data gathered at a single moni- toring station. The question is conceptually troubling to begin with be- cause the variability of water quality is large and continuous in both space and time. In practice, moreover, the de facto valid region for monitoring stations is extremely large. Given the spatially detailed treatment of rivers and streams in the NHD, however, most states would need to gather data from more than a thousand stations per year to main- tain an average “monitoring ratio” of 100 km per station (assuming the NHD approximately describes state waters). This distance is clearly greater than the valid region for monitoring stations on most surface wa- ters, especially because most of the channel length in state waters is con- tributed by relatively small streams (e.g., drainage areas less than 100 km2) where water quality conditions may vary greatly over short dis- tances. Thus, a substantial portion of state waters would appear to be located outside of the valid monitoring region for a state monitoring pro- gram of 1,000 stations. These waters are either left out of the decision process and are deemed not impaired by default, or they are included in the decision process with higher error rates. One solution to this problem is to avoid the concept of a valid region for individual monitoring stations entirely and replace it with an ap- proach in which monitoring data are used to develop statistical models of water quality in state waters. Water quality conditions at monitoring sites can be statistically related to known factors that cause impairment in watersheds (the size and location of stressors, for example), thus ena- bling estimates of water quality conditions at other unmonitored loca- tions. As discussed later, this approach may also benefit the listing proc- ess. 

44 Assessing the TMDL Approach to Water Quality Management Conclusions and Recommendations 1. Each state should develop a catalogue of waterbodies based on the National Hydrography Dataset for the purposes of defining state waters and designing sampling and assessment programs. 2. States should attempt to move away from the concept of a re- gion of validity of individual monitoring stations and instead con- sider a statistical modeling approach to assessing the condition of waters. This approach would combine monitoring data with estimates of water quality based on statistical models. DESIRABLE CRITERIA This section considers the desired features of chemical and biological criteria as surrogates for designated use. For listing and delisting pur- poses, numeric and measurable criteria should be logically derived from the designated use statement. Ideally, appropriate designated uses and associated criteria are assigned to each waterbody prior to an assessment. Realistically, the cost and effort involved in categorizing every water- body in advance of an assessment may be prohibitive, and many states’ programs for setting appropriate use designation are continuing efforts. As is noted in Chapter 5, it is advisable to conduct a site-specific review to refine the standard once a waterbody is listed and before a TMDL is initiated. One desired feature of a criterion is that it must be measurable with available monitoring methods. Unfortunately, federal guidelines for wa- ter quality assessment (EPA, 1994) do not assure this feature. In many cases there may be a discrepancy between the formulation of water qual- ity criteria and the frequency with which water quality data are gathered. A criterion may not be a single number, but instead may be repre- sented as a frequency, duration, and magnitude. In the context of a pol- lutant, the magnitude refers to how much of the pollutant can be allowed in the water while still achieving the designated use. The magnitude can be chosen to protect against either acute or chronic effects of a pollutant. Duration refers to the period of time over which measurements of the pollutant are considered. Pollutant levels may be averaged over some number of hours or days to determine that amount of the pollutant that can be present without a loss of the designated use. The allowable fre- 

Waterbody Assessment: Listing and Delisting 45 quency at which the criterion can be violated (called an excursion) with- out a loss of the designated use also must be considered. Thus, in the case of a trout fishery, the criterion might specify a minimum DO (or maximum chlorophyll a) that can be realized for a period of time and the number of times this number can be violated before there is demonstra- ble harm to the designated use. It should be noted that these numbers are pollutant-specific, and they might vary with season depending on, for example, fish life-stage. Establishing these three dimensions of the criterion is crucial for successfully developing water quality standards8. Currently, there are many cases where there are insufficient data collected in one or more of these three dimensions to evaluate attainment of water quality criteria. In addition, some standards are virtually impossible to comply with, espe- cially when the frequency of allowable excursions is zero (called “no- exceedance” standards). Box 3-4 provides three examples of criteria that are either unmeasurable given current monitoring protocols or are ex- ceedingly difficult to meet and thus constitute an intractable problem for the TMDL program. Careful consideration of the three dimensions of the criterion is also critical to the development of appropriate TMDLs. In the law, the letter “d” in TMDL refers to a daily load, which has been interpreted literally in some legal cases. However, for many pollutants, the load determined over a longer time period (e.g., a season or year) is more relevant to securing the designated use. Examples of this are nutri- ent and sediment criteria, where the duration component of the criterion is generally not stated as “daily.” A second desirable feature is that the measured criterion must be logically derived from the qualitative statement of the designated use. The closer the criterion is in the causal chain (Figure 2-1), the easier it is to make that connection. This has led to increased interest in biocriteria, particularly numeric measures of fish, benthic invertebrate, algal, and diatom assemblages. Recommendations to adopt biocriteria are often made because biocriteria integrate the effects of multiple stressors over time and space, thus minimizing the need for a large number of samples (Karr, 2000). A second advantage of using biocriteria is that, unlike chemical criteria, they are designed to be specific to certain regions and 8 Specifying the magnitude, frequency, and duration is critical for chemical cri- teria, but may not be necessary for certain biological criteria. For example, the fecal coliform standard is best defined with all three components. On the other hand, many biocriteria such as IBI are well defined by a single number because they integrate biological, chemical, and physical effects over time. 

46 Assessing the TMDL Approach to Water Quality Management BOX 3-4 Problems Associated with Standards Unmeasurable Standards By definition, the TMDL program requires that waterbodies meet water quality criteria daily, interpreted by some as meaning that the sampling frequency must be daily. This requires that a complete time series of grab or composite samples be taken daily without an interruption over a period of a minimum of three years. As one might expect, such time series of water quality data are almost never available for waterbody assessment (with the exception of the continuous monitoring for a few parameters such as DO or temperature). Samples are generally taken monthly for common parameters and annually or less often for some toxic chemicals that require expensive laboratory analytical methodology. Sediment sampling is done infrequently, perhaps once in a period of several years. Similarly, the frequency/duration components of water quality criteria for contact recreation are generally infeasible to measure. Many states use fecal coliform count as an indicator for the contact recreation. The standards are usually compared to the geometric mean of at least five samples taken over 30 days. This standard is not defined in terms of allowable excursions; thus, there is no frequency component. With the exception of waterbodies used for water supply, monitoring data are rarely collected often enough to comply with such a standard. No-Exceedance Standard Many states require that a numeric standard be maintained at all times, which implies that all monitored values of a parameter should be below the criterion. Such a limitation is a statistical impossibility because there is always a chance—albeit remote—that a water parameter may reach a high but statistically possible value exceeding an established standard. In addition, this requirement would seem to provide an incen- tive to sample as little as possible in order to reduce the chance of col- conditions. For example, a swamp forest will typically violate DO crite- ria, and waterbodies in mountain areas with heavy metal-bearing rocks may violate heavy metal criteria. Biocriteria that are regionally relevant would not show those conditions as violations. Fecal coliform counts and algal community parameters such as chlo- rophyll a are a type of biocriteria, but they are not comprehensive meas- ures of waterbody condition. To make bioassessment more comprehen- 

Waterbody Assessment: Listing and Delisting 47 lecting a sample that is in exceedance. For example, it is possible that if nine samples are taken over a period of three years, none of the sam- ples would, by chance, result in an excursion. If 100 samples are taken in the same period, a few (e.g., five or less) may exceed the standard. The former sampling scheme would indicate that the waterbody is in compliance while the other would not. Stream concentrations represent statistical time series for which only infinitesimally large values of a stan- dard would have a 100 percent statistical probability of not ever being exceeded. Flow Restriction Standards To make “no-exceedance” standards easier to comply with, EPA (1992) and many states incorporated a flow restriction into the standards. Thus, the standards must be main-tained at all times except at flows that are less than some specified low flow value (one example is given be- low). Unfortunately, except for the “harmonic mean flow” (Singh and Ramamurthy, 1991), none of the critical low flows specified by EPA allow consideration of wet weather discharges (Novotny, 1999). Thus, under wet weather flows, the “no-exceedance” criterion is in effect. This ig- nores the fact that measured water quality parameters are naturally vari- able. One type of flow restriction standard is based on hydrologically based design flows. To protect against acute effects, such water quality criteria must be met at all times except during the lowest daily flow occur- ring once every 10 years (referred to as 1Q10). To protect against chronic effects, water quality criteria must be met at all times except during the lowest flow occurring once every 10 years averaged over a 7- consecutive-day period (7Q10). This approach assumes that concentra- tions of pollutants of concern are decreasing as flows increase—likely to be true for the case of a continuous year-round discharge from a point source, but not for nonpoint sources. It should be noted that these de- sign flows have “interim” status and were not recommended for general application with water quality standards. In addition, hydrologically based design flows vary from state to state. sive, index systems have been developed that focus on characteristics of the biota expected in the particular region where the waterbody is lo- cated, including desired fish species and other associated organisms (Box 3-5). The scientific community measures integrity by describing the bio- logical condition of waterbodies that, as much as possible, have not been altered by human activity. When “pristine” or “minimally disturbed” 

48 Assessing the TMDL Approach to Water Quality Management BOX 3-5 Index Systems for Bioassessment During the past two decades, biological assessmentevaluating human-caused biotic changes apart from those occurring naturallyhas become a part of water managers’ tool kits. Two major approaches to ambient biological monitoring are used—the river invertebrate prediction and classification system (RIVPACS) and the multimetric index of bio- logical integrity (IBI). Although their conceptual and analytical details differ, both RIVPACS and IBI (1) focus on biological endpoints to define waterbody condition, (2) use a concept of a regionally relevant reference condition as a benchmark, (3) organize sites into classes with similar environmental characteristics, (4) assess change and degradation caused by human effects, (5) require standardized sampling, laboratory, and analytical methods, (6) score sites numerically to reflect site condi- tion, (7) define “bands,” or condition classes, representing waterbody condition, and (8) furnish needed information for diverse management decisions (Karr and Chu, 2000). RIVPACS was developed in England (Wright et al., 1989, 1997) with clones available for use in Australia (Norris et al., 1995) and Maine (Da- vies and Tsomides, 1997). IBI was developed in the United States (Karr, 1981; Karr et al., 1986; Karr and Chu, 1999) with clones applied by state and federal agencies (Ohio EPA, 1988; Davis et al., 1996; Barbour et al., 1999) and abroad (Hughes and Oberdorff, 1999). Although applications of RIVPACS are historically limited to invertebrates in rivers, IBI applica- tions have been developed for diverse taxonomic groups and waterbody types. For example, a multimetric index (RFAI, reservoir fish assessment index) has been developed as a component of Tennessee Valley Author- ity’s (TVA) “vital signs” monitoring program to assess fishery manage- ment success in reservoirs (Jennings et al., 1995; McDonough and Hickman, 1999). As a general example, consider a minimally disturbed Pacific North- west stream supporting self-sustaining populations of salmon and asso- ciated assemblages of invertebrates. With urban development, salmon decline and cutthroat trout become relatively more abundant, and certain invertebrate taxa (e.g., stoneflies) are reduced or eliminated. Tiered beneficial uses could in this case differentiate between streams support- ing salmon vs. cutthroat trout, using an index based on the invertebrate assemblage as the biocriterion. Recent work in these streams suggests that a benthic index of biological integrity (B-IBI) of about 35 is a mini- mum required to maintain a healthy salmon population (Karr, 1998). If the IBI drops below 20 because of continued development, even the cutthroat trout will eventually disappear. 

Waterbody Assessment: Listing and Delisting 49 sites are used to define integrity, any site that has been altered by human actions must, by definition, lack integrity because its biota have changed in response to the actions of humans. For obvious reasons, reservoirs, farm ponds, and other waterbodies “created” by human actions cannot be assessed using this standard. However, it does not follow that a waterbody lacking integrity is im- paired or that restoring biological integrity is either possible or desirable. A waterbody that is described as lacking “biological integrity” should not be assumed to be in a less-than-desirable state. Rather, when a bio- assessment finds that a waterbody diverges from integrity, there must be a social decision about whether that divergence is acceptable. In short, “The biota of minimally disturbed sites—those with integrity— provides a benchmark, a standard by which others are measured. The protection of that standard, or something very close to it, is likely to be the goalthe end toward which effort is directedin relatively few places (e.g., national parks). The modern reality is that we are not able to preserve all areas in this benchmark condition. For example, restoring salmon to every Pacific Northwest stream is not re- alistic, yet a restoration goal that includes viable populations of cut- throat trout may be reasonable even in many urban or suburban streams. (Karr, 2000) Measures of biological condition (e.g., IBI) inform society of the status of a water resource. But society must decide the desired designated use and then determine what level on the index numeric scale is, with rea- sonable certainty, likely to protect that designated use. Recently, the EPA Office of Water has convened a working group of states and other supporting institutions to better define the gradient of biological condition from pristine to highly degraded and link this with operational measures such as numeric biocriteria in a manner that will ensure consistency across state programs. This is referred to as tiered aquatic life uses and is expressed as a biocondition axis. Examples of this framework already exist in Maine, Ohio, and Vermont. The expec- tation is that as states develop a more detailed system of tiered desig- nated uses, they will also develop measurable biocriteria logically tied to those uses. 

50 Assessing the TMDL Approach to Water Quality Management Conclusions and Recommendations 1. All chemical criteria and some biological criteria should be defined in terms of magnitude, frequency, and duration. Each of these three components is pollutant-specific and may vary with season. The frequency component should be expressed in terms of a number of allowed excursions in a specified period (return period) and not in terms of the low flow or an absolute “never to be exceeded” limit. The re- quirement of “no exceedances” for many water quality criteria is not achievable given natural variability alone, much less with the variability associated with discharges from point and nonpoint sources. 2. Water quality standards must be measurable by reasonably obtainable monitoring data. In many states, there is a fundamental dis- crepancy between the criteria that have been chosen to determine whether a waterbody is achieving its designated use and the frequency with which water quality data are collected. 3. Biological criteria should be used in conjunction with physical and chemical criteria to determine whether a waterbody is meeting its designated use. Biocriteria are more closely related to designated uses, they can be defined and measured, and they integrate the effects of multiple stressors over time and space. LISTING AND DELISTING IN A DATA-LIMITED ENVIRONMENT As discussed at the beginning of this chapter, states are confronted with lengthy lists of impaired waters requiring TMDLs, many of which were judged against inadequate standards or were not fully assessed as part of a comprehensive ambient monitoring program. This section pro- poses a mechanism for managing the large number of waters requiring attention by dividing the listing process into multiple smaller steps, as shown in Figure 3-1. Figure 3-1 illustrates a framework for water quality management that is more detailed than the conceptualized steps of the TMDL process shown in Figure 1-1. Figure 3-1 begins with the identification of all wa- ters to be assessed and the determination of appropriate water quality standards as in the current TMDL program. Following this, however, 

Waterbody Assessment: Listing and Delisting 51 All Waters Determine Designated Use/ Standard Screening Assessment “Preliminary” List Full Review Use/ Assessment Standard “Action” List (303d) TMDL Planning Adaptive Implementation FIGURE 3-1 Framework for water quality management. 

52 Assessing the TMDL Approach to Water Quality Management waters to be assessed would next go through an initial screening assess- ment. This involves comparing available, and often limited, data on wa- ter quality conditions with the existing applicable water quality criterion. If based on this initial screening assessment the waterbody is considered a candidate for impairment, it is advanced to the “preliminary” list for further consideration. It should be relatively easy to get on the prelimi- nary list, the consequences of which include additional and immediate investigation to determine the nature and reality of a suspected problem. The term “preliminary” indicates that waterbodies on this list may later be placed on an action list, but they may also be declared unimpaired. Such a preliminary list has been suggested or employed in some states (e.g., Florida). Those waterbodies placed on the preliminary list are the object of a more complete assessment that would involve additional monitoring and appropriate analysis of new data to reduce the uncertainty about their condition. If the decision from the full assessment is that the waterbody is impaired, then it moves to an “action list.” One might think of the ac- tion list as the state’s impaired waters (303d) list. The word “impaired” is a term of art. Impaired waters under Section 303d are analogous to “water quality limited segment(s),” as defined in the federal regulations (40 CFR Section 130.2(j)). The consequence of advancing to the action list is that additional resources are needed to either review and update the existing standard or complete a TMDL. (For those cases in which the existing criteria are not appropriate to a waterbody, Figure 3-1 allows for review of the water quality standard for that waterbody. The process for completing that review—use attainability analysis—is discussed in Chapter 5.) The organizing concept in this idealized process is continuous and concurrent progress toward improved monitoring and listing decisions. The process moves forward from a position of limited information to more information; from uncertainty to more certainty; and from inaction to progressively larger and possibly more costly actions. Were EPA to endorse the idealized process represented in Figure 3-1, the listing proc- ess would be improved. For example, at the current time, there are thou- sands of waters on state 303d lists that were not placed there using ade- quate data or information. Waters in this category should be moved back to the preliminary list, represented by the dashed return arrow in Figure 3-1, to allow a more complete evaluation to be made. 

Waterbody Assessment: Listing and Delisting 53 Creating the Preliminary List Determining whether there should be some minimum threshold of data available when evaluating waterbodies for attainment of water qual- ity standards is an issue of great concern to states. On the one hand, many call for using only the “best science” in making listing decisions, while others fear many impaired waters will not be identified in the wait for additional data. The existence of a preliminary list addresses these concerns by focusing attention on waters suspected to be impaired with- out imposing on stakeholders and the agencies the consequences of TMDL development, until additional information is developed and evaluated. In many cases, biological and limited water quality surveys along with an inventory of existing sources of pollution may provide adequate information for a screening assessment of the waterbody. Evaluated data are also an important source of information for determining if a water- body should be placed on the preliminary list. Evaluated data may take many forms (e.g., data older than a certain age, beach closures based on fixed rainfall thresholds, visual observations, and statistical inferences from small data sets) and have been described differently from state to state9. In contrast, monitored data are viewed as being more comprehen- sive, typically using data less than five years old, and may include a wide array of direct measurements of water quality, including physical, chemi- cal, or biological measures. Use of evaluated data has been controversial in water quality assessments under the Clean Water Act. The contro- versy would be lessened if the use of evaluated data were limited to placing waters on the preliminary list. The quality of the data used to list waterbodies as impaired is fre- quently a concern. Beyond the normal data entry, sampling, and labora- tory errors, states must determine the reliability of the data coming from a wide range of sources (especially for evaluated data). Some states have responded to this uncertainty by strictly limiting the data used in making assessments to those collected by the state’s lead environmental agency or some other select group of data providers (such as USGS). To over- 9 Evaluated data and/or information provides an indirect appraisal of water qual- ity through such sources as information on historical adjacent land uses, aquatic and riparian health and habitat, location of sources, results from predictive mod- eling using input variables, and some surveys of fish and wildlife. Monitored data refers to direct measurements of water quality, including sediment meas- urements, bioassessments, and some fish tissue analyses (EPA, 1998, 2000). 

54 Assessing the TMDL Approach to Water Quality Management come this uncertainty, and thereby expand the universe of reliable data, some states have required that associated meta data10 be provided and entered into a central data repository (such as STORET). Narrative criteria might also play a significant role in determining whether a waterbody should be placed on the preliminary list. Many water quality standards are characterized only by narrative criteria that express the desired target but do not allow comparison to a numeric value. For example, a typical narrative criterion for nutrients (nitrogen and phosphorus) in inland waters is “concentrations should be limited to the extent necessary to prevent nuisance growths of algae, weeds, and slimes” (as in New York State). Currently, violations based on interpre- tation of a narrative criterion may be a basis for placing a waterbody on the 303d list, even though such an evaluation is done without a numeric value of the criterion. EPA and the states have worked together over the last ten years to develop translators that will convert narrative standards to numeric criteria or guidance values (EPA, 1999b,c; NRC, 2000). While further progress is made in developing such translators, violations of narrative standards should be used to place waterbodies on the pre- liminary list. The approaches to creating a preliminary list will vary from state to state. For example, in Florida, data and information used to place waters on the preliminary list have to meet certain basic QA/QC requirements as well as limited data sufficiency tests. Minimum sample sizes and confi- dence levels have been established, and both chemical and biological data are considered. States will have to decide upon and develop criteria for defining data sufficiency and analytical procedures for placing water- bodies on the preliminary list and the action list. EPA might be expected to assist in this process. Moving Off the Preliminary List Waters on the preliminary list should receive special monitoring at- tention. Movement from the preliminary list will be either back to the list of all waters or onto the action (303d) list. Movement off the pre- liminary list will demand a more analytically structured evaluation than 10 Meta data is information about data and its usage, such as (1) what it is about, (2) where it is to be found, (3) how much it costs, (4) who can access it, (5) in what format it is available, (6) what the quality of the data is for a specified pur- pose, and (7) what spatial location and time period it covers. 

Waterbody Assessment: Listing and Delisting 55 was required for getting on the list. Each state should develop statistical procedures appropriate for testing attainment of each criterion. Sampling design, sample size, and QA/QC assurances for monitoring data would be defined, as would the appropriate tools for data analysis. If the data evaluated by the appropriate procedure indicate that there is no impair- ment, then delisting would follow. Delisting depends on analyses of sampling data and not on the implementation of a TMDL plan, although such a plan may be required to meet the criterion. The process represented in Figure 3-1 is designed to improve the ac- curacy of the listing process. Placement of a waterbody on the prelimi- nary list can serve as an indication to stakeholders that action should be taken soon to achieve water quality standards in order to avoid the costs associated with TMDL development. Because of the consequences of movement to the action list, there may be an incentive to keep waters on the preliminary list indefinitely. This incentive can be eliminated by re- quiring that a waterbody be automatically placed on the action (303d) list at the end of the next rotating basin cycle if additional analyses have not been undertaken. Such a requirement also may provide an incentive for point and nonpoint pollutant sources to contribute to the monitoring pro- gram in order to (potentially) avoid the consequences of a 303d listing. Conclusions and Recommendations 1. EPA should approve the use of both a preliminary list and an action list instead of one 303d list. The two-list process would reduce the uncertainty that often accompanies a listing decision and would pro- vide flexibility to the TMDL program. 2. If some waters on the current 303d list would be more appro- priately catalogued on the preliminary list, EPA should allow states to move those waterbodies from the current 303d list to the prelimi- nary list. If no legal mechanism exists to bring this about, Congress should create one. Many waters now on state 303d lists were placed there without the benefit of adequate data or waterbody assessment. These potentially erroneous listings contribute to a very large backlog of TMDL segments and foster the perception of a problem that is larger than it may actually be. 3. States should be allowed the flexibility to delist a waterbody without having to complete a TMDL if additional data or new in- 

56 Assessing the TMDL Approach to Water Quality Management formation providing evidence of attainment of the water quality standard becomes available. 4. No waterbody should remain on the preliminary list for more than one rotating basin cycle. If the waterbody has not been removed from the preliminary list at the end of a rotating basin cycle, it should automatically be placed on the 303d list, unless EPA approves an ex- emption from such a requirement on a waterbody-by-waterbody basis. Criteria for granting exemptions could be developed by EPA. 5. To increase the reliability of the data used in listing water- bodies, EPA should require some limited amount of meta data for data submitted to STORET. DATA EVALUATION FOR THE LISTING AND DELISTING PROCESS Given finite monitoring resources, it is obvious that the number of sampling stations included in the state program will ultimately limit the number of water quality measurements that can be made at each station. Thus, in addition to the problem of defining state waters and designing the monitoring network to assess those waters, fundamental statistical issues arise concerning how to interpret limited data from individual sampling stations. Statistical inference procedures must be used on the sample data to test hypotheses about whether the actual condition in the waterbody meets the criterion. Thus, water quality assessment is a hy- pothesis-testing procedure. A statistical analysis of sample data for determining whether a wa- terbody is meeting a criterion requires the definition of a null hypothesis; for listing a waterbody, the null hypothesis would be that the water is not impaired11. The analysis is prone to the possibility of both Type I error (a false conclusion that an unimpaired water is impaired) and Type II error (a false conclusion that an impaired water is not impaired). Differ- ent statistical analyses are needed depending on whether chemical or biological criteria are being assessed. 11 For delisting, the null hypothesis might be that the water is impaired. 

Waterbody Assessment: Listing and Delisting 57 Statistical Approaches for Chemical Parameters If chemical criteria—carefully designed to account for magnitude, frequency, and duration—are expected to be met, instantaneous meas- urements would be needed to determine compliance. Under current practice, however, even when states conduct frequent monitoring, sample sizes are limited, and so the possibility for false positive errors (Type I) and false negative errors (Type II) remains. As sample sizes increase, error rates can be better managed. For placement on the preliminary list, a small sample size may be acceptable. However, placement on the ac- tion list would require an increase in the number of sample points used in order to reduce the uncertainty in the listing and delisting decisions. The committee does not recommend any particular statistical method for analyzing monitoring data and for listing waters. However, one pos- sibility is that the binomial hypothesis test could be required as a mini- mum and practical first step (Smith et al., 2001). The binomial method is not a significant departure from the current approach—called the raw score approach—in which the listing process treats all sample observa- tions as binary values that either exceed the criterion or do not, and the binomial method has some important advantages. For example, one limitation of the raw score approach is that it does not account for the total number of measurements made. Clearly, 1 out of 6 measurements above the criterion is a weaker case for impairment than is 6 out of 36. The binomial hypothesis test allows one to take sample size into account. By using a statistical procedure, sample sizes can be selected and one can explicitly control and make trade-offs between error rates (see Smith et al., 2001, and Gibbons, in press, for guidance on managing the risk of false positive and false negative errors)12. Several states, including Florida and Virginia, are considering or are already using the binomial hypothesis test to list impaired waters. Detailed examples of how to ap- 12 The choice of a Type I error rate is based on the assessors willingness to falsely categorize a waterbody. It also is the case that, for any sample size, the Type II error rate decreases as the acceptable Type I error rate increases. The willingness to make either kind of mistake will depend on the consequences of the resulting actions (more monitoring, costs to do a TMDL plan, costs to im- plement controls, possible health risk) and who bears the cost (public budget, private parties, etc.). The magnitude and burden of a Type I versus Type II error depend on the statement of the null hypothesis and on the sample size. When choosing a Type I error rate, the assessor may want to explicitly consider these determinants of error rates. 

58 Assessing the TMDL Approach to Water Quality Management ply this test are beyond the scope of this document, but can be found in Smith et al. (2001) and the proposed Chapter 62-303 of the Florida Ad- ministrative Code13. Whether the binomial or the raw score approach is used, there must be a decision on an acceptable frequency of violation for the numeric criterion, which can range from 0 percent of the time to some positive number. Under the current EPA approach, 10 percent of the sample measurements of a given pollutant made at a station may exceed the ap- plicable criterion without having to list the surrounding waterbody. The choice of 10 percent is meant to allow for uncertainty in the decision process. Unfortunately, simply setting an upper bound on the percentage of measurements at a station that may violate a standard provides insuffi- cient information to properly deal with the uncertainty concerning im- pairment. The choice of acceptable frequency of violation is also supposed to be related to whether the designated use will be compromised, which is clearly dependent on the pollutant and on waterbody characteristics such as flow rate. A determination of 10 percent cannot be expected to apply to all water quality situations. In fact, it is inconsistent with federal wa- ter quality criteria for toxics that specify allowable violation frequencies of either one day in three years, four consecutive days in three years, or 30 consecutive days in three years (which are all less than 10 percent). Embedded in the EPA raw score approach is an implication that 10 per- cent is an acceptable violation rate, which it may not be in certain cir- cumstances. Both the raw score and binomial approaches require the analyst to “throw away” some of the information found in collected data. For ex- ample, if the criterion is 1.0, measurements of 1.1 and 10 are given equal importance, and both are treated simply as exceeding the standard. Thus, a potentially large amount of information about the likelihood of im- pairment is simply discarded. (The standard deviation can be used to set priorities for TMDL development or other restoration activities.) There are other approaches that are more effective at extracting information from a single monitoring sample, thereby reducing the number of sam- ples needed to make a decision with the same level of statistical confi- 13 This proposed rule chapter was approved for adoption by the Florida Depart- ment of Environmental Protection’s Environmental Regulation Commission on April 26, 2001, but has not been officially filed for adoption by the Department because of a pending rule challenge before the Division of Administrative Hearings. 

Waterbody Assessment: Listing and Delisting 59 dence. For example, Gibbons (in press) suggests testing the data for normality or log normality and then examining the confidence intervals surrounding the estimated 90th percentile of the chosen distribution. When the data are neither normal nor lognormal, or when more than 50 percent of the observations are censored (below the detection limit), Gibbons suggests constructing a nonparametric confidence limit based on the binomial distribution of ranked data. Another approach that uses all the data to make a decision is “acceptance sampling by variables” (Duncan, 1974). In general, alternative statistical approaches transform questions about the proportion of samples that exceed a standard into questions about the center (or another parameter) of a continuous distri- bution. It should be noted that new approaches will bring new analytical requirements that must be taken into consideration. For example, if there is a requirement to specify a distribution, sufficient data must be avail- able. In some cases, data from other similar sites may be needed to give an overall assessment of distribution type. Finally, as more powerful statistical procedures are used, water quality assessors will need to un- derstand how to run the tests and also how to state hypotheses that clearly relate to the water quality criterion. Statistical Approaches for Biological Parameters Error bands exist with any sampled data, including bioassessment re- sults. Thus, bioassessment procedures must also be designed to be sta- tistically sound. The utility of any measure of stream condition depends on how accurately the original sample represents the condition in the stream—that is, how successful it is in avoiding statistical “bias.” Proto- cols to for making such measurements are established in the technical literature (Karr and Chu, 1999) as well as in guidance manuals produced by EPA (Barbour et al., 1996, 1999; EPA, 1998a; Gibson et al., 2000). There are three principal ways variability is dealt with in the process of deriving and using biocriteria (Yoder and Rankin, 1995). First, vari- ability is compressed through the use of multimetric evaluation mecha- nisms such as IBI. Reference data for each metric are compressed into discrete scoring ranges (i.e., 5, 3, and 1). Second, variability is stratified via tiered uses, ecoregions, stream size categories (headwaters, wadable, boatable), and method of calibrating each metric (i.e., vectoring expecta- tions by stream size). Third, variability is controlled through standard- ized operating procedures, data quality objectives (i.e., level of taxon- omy), index sampling periods (to control for seasonal effects), replication 

60 Assessing the TMDL Approach to Water Quality Management of sampling, and training (Yoder and Rankin, 1995). One can, for exam- ple, avoid seasonal variation by carefully defining index sampling peri- ods or variation among microhabitats by sampling the most representa- tive microhabitat (Karr and Chu, 1999). Box 3-6 presents results of sev- eral studies in which the error around biological parameters was as- sessed. BOX 3-6 Understanding Sources of Variability in Bioassessment Sources of error evaluated in one study of biological monitoring data from New England lakes (Karr and Chu, 1999) included three types of variance: interlake variability (differences among lakes); intralake vari- ability (variability associated with sampling different sites within a lake as decided by the field crew), and lab error (error related to subsample work in the lab). The interlake variability was the effect of interest, and the goal was to determine if that source of variability was dominant. Distri- bution of variance varied as a function of biological metric selected. Those measures with reduced variance except for the context of interest (e.g., interlake variability) were selected for inclusion in IBI to increase the probability of detecting and understanding the pattern of interest. Two other studies involved an examination not of the individual met- rics, but of the overall IBI (i.e., after individual metrics were tested and integrated into an IBI). For Puget Sound streams, 9 percent of variation came from differences within streams and 91 percent was variability across streams (reported in Karr and Chu, 1999, Fig. 35). For a study in Grand Teton National Park, streams were grouped in classes reflecting different amounts of human activity in their watersheds. In this case, 89 percent of the variance came from differences among the groups, and 11 percent came from differences among members of the same group (re- ported in Karr and Chu, 1999). In all these cases, the goal was to find ways of measuring that em- phasize differences among watersheds with differing human influences, while keeping other sources of variation small. Success in these exam- ples was based on the development of an earlier understanding of sources of variation and then establishing sampling protocols that avoid other irrelevant sources of variation (such as variation stemming from the differing abilities of personnel to select and use methods). If these sources of variation are controlled for, then the study can emphasize the kind of variation that is of primary interest (e.g., human influence gradi- ents). 

Waterbody Assessment: Listing and Delisting 61 Conclusions and Recommendations 1. EPA should endorse statistical approaches to proper moni- toring design, data analysis, and impairment assessment. For chemi- cal parameters, these might include the binomial hypothesis test or other statistical approaches that can more effectively make use of the data col- lected to determine water quality impairment than does the raw score approach. For biological parameters, these might focus on improvement of sampling designs, more careful identification of the components of biology used as indicators, and analytical procedures that explore bio- logical data as well as integrate biological information with other rele- vant data. 2. States should be required to report the statistical properties of the sample data analyses used to make listing determinations. Error rates, confidence limits, or other means of conveying uncertainty should be presented along with the rationale for a decision to list or delist a wa- terbody. USE OF MODELS IN THE LISTING PROCESS As stated in EPA guidance documents as well as the Federal Advi- sory Committee Act (FACA) report (EPA, 1998b), monitoring data are the preferred form of information for identifying impaired waters. Model predictions might be used in addition to or instead of monitoring data for two reasons: (1) modeling could be feasible in some situations where monitoring is not, and (2) integrated monitoring and modeling systems could provide better information than monitoring alone for the same total cost. EPA guidance and the FACA report explicitly recognize the obvious practicality of the first reason, but largely ignore the poten- tial importance of the second. This section considers some of the ways in which modeling might be used as a complement to monitoring and points out some limitations of modeling in informing the listing process. Often, in attempting to estimate the frequency of violation of a stan- dard, the number of pollutant concentration measurements made in a waterbody is so small that it is difficult to avoid false negative error with the desired level of confidence. One way in which a simple statistical model may assist in interpreting monitoring data in such cases is by in- troducing a variable to the analysis that is correlated with pollutant con- centration. One common correlate of many water quality time series is 

62 Assessing the TMDL Approach to Water Quality Management stream flow, which is measured continuously at many monitoring sta- tions, including nearly all USGS stations. The statistical methods for taking advantage of correlated stream flow data are called record exten- sion techniques, several of which have been described and compared by Hirsch (1982). By modeling pollutant concentration as a function of streamflow and using the resulting model to estimate a denser concentra- tion time series, a better estimate of the frequency distribution of pollut- ant concentration may be obtained. The predicted concentration time series then may be tested for violation frequency using either the bino- mial approach (see above) or the quantile approach. The value of this modeling approach over using pollutant data alone is directly dependent on the level of correlation that exists between the pollutant concentration and stream flow. Further discussion of the specific extension technique called MOVE (Maintenance of Variance–Extension) appears in Helsel and Hirsch (1991). The EPA guidance on 303d listing suggests that a simple, but useful, modeling approach that may be used in the absence of monitoring data is “dilution calculations,” in which the rate of pollutant loading from point sources in a waterbody (recorded as kg per day in NPDES permits, for example) is divided by the stream flow distribution to give a set of esti- mated pollutant concentrations that may be compared to the state stan- dard. Simple dilution calculations assume conservative movement of pollutants through a watershed and ignore the fact that for most pollut- ants some loss of mass occurs during transport due to a variety of proc- esses including evaporation, settling, or biochemical transformation (see, for example, Novotny and Olem, 1994). Thus, the use of dilution calcu- lations will tend to bias the decision process toward false positive con- clusions. Lacking a clear rationale for such a bias, a better approach would be to include a best estimate of the effects of loss processes in the dilution model. Section 303d and related guidance from EPA emphasize the impor- tance of searching for information on waterbodies that are suspected of violating water quality standards, which is understandable given the de- sire to limit the number of sites sampled and hence the cost of monitor- ing. Targeted monitoring will often increase the efficiency of the as- sessment process (i.e., reduce the total number of decision errors), but may have somewhat hidden effects on the balance of false positive and false negative errors. Targeted monitoring represents the informal use of a prior probability distribution on impairment to guide monitoring to- ward sites located in a particular region of the distribution. One of the 

Waterbody Assessment: Listing and Delisting 63 most potentially valuable uses of modeling in relation to 303d listing would be to formalize the use of prior information on impairment prob- ability in order to better organize the decision process. That is, modeling techniques such as SPARROW (Smith et al., 1997) could be used to es- timate preliminary impairment distributions for all waterbodies in the state. These distributions would then be used to guide monitoring and control the rates of false positive and false negative error either through Bayesian or other methods of interpreting monitoring data. Limited monitoring resources generally could be focused on the sites where im- pairment was most uncertain (i.e., where the estimated probability of im- pairment was neither very high nor very low), potentially improving the efficiency of monitoring. Sites at the extremes of the impairment distri- butions (i.e., extremely likely or unlikely to be impaired) would be less frequently monitored. Decisions for placing waters on a preliminary list might be made primarily on the basis of such modeling. (Formal place- ment of a waterbody on the 303d list would require additional monitor- ing.) Conclusions and Recommendations 1. Models that can fill gaps in data have the potential to generate information that will increase the efficiency of monitoring and thus increase the accuracy of the preliminary listing process. For example, regression analyses that correlate pollutant concentration with some more easily measurable factor could be used to extend monitoring data for preliminary listing purposes. Models can also be used in a Bayesian framework to determine preliminary probability distributions of impair- ment that can help direct monitoring efforts and reduce the quantity of monitoring data needed for making listing decisions at a given level of reliability. REFERENCES Barbour, M. T., J. B. Stribling, J. Gerritsen, and J. R. Karr. 1996. Biological Criteria: Technical Guidance for Streams and Small Rivers. Revised Edi- tion. EPA 822-B-96-001. Washington, DC: EPA Office of Water. Barbour, M. T., J. Gerritsen, B. D. Snyder, and J. B. Stribling. 1999. Rapid Bioassessment Protocols for Use in Streams and Wadeable Rivers: Peri- 

64 Assessing the TMDL Approach to Water Quality Management phyton, Benthic Macroinvertebrates and Fish, Second Edition. EPA 841-B- 99-002. Washington, DC: EPA Office of Water. Davis, W. S., B. D. Snyder, J. B. Stribling, and C. Stoughton. 1996. Summary of State Biological Assessment Programs for Streams and Rivers. EPA 230-R-96-007. Washington, DC: EPA Office of Policy, Planning, and Evaluation. DeWald, T., R. Horn, R. Greenspun, P. Taylor, L. Manning, and A. Montalbano. 1985. STORET Reach Retrieval Documentation. Washington, DC: EPA. Environmental Protection Agency (EPA). 1994. Water Quality Standards Handbook. Second Edition. EPA 823-B-94-005a. Washington, DC: EPA Office of Water. EPA. 1995a. Environmental indicators of water quality in the United States. EPA 841-R-96-002. Washington, DC: Office of Policy, Planning, and Evaluation. EPA. 1995b. A conceptual framework to support development and use of envi- ronmental information in decision-making. EPA 239-R-95-012. Wash- ington, DC: Office of Policy, Planning, and Evaluation. EPA. 1998a. Lake and Reservoir Bioassessment and Biocriteria: Technical Guidance Document. EPA 841-B-98-007. Washington, DC: EPA Office of Water. EPA. 1998b. Report of the FACA Committee on the TMDL Program. EPA 100-R-98-006. Washington, DC: EPA Office of the Administrator. EPA. 1999a. Draft Guidance for Water Quality-based Decisions: The TMDL Process (Second Edition). Washington, DC: EPA Office of Water. EPA. 1999b. Protocol for Developing Sediment TMDLs. First Edition. EPA 841-B-99-004. Washington, DC: EPA Office of Water. EPA. 1999c. Protocol for Developing Nutrient TMDLs. First Edition. EPA 841-B-99-007. Washington, DC: EPA Office of Water. General Accounting Office (GAO). 2000. Water Quality-Key EPA and State Decisions Limited by Inconsistent and Incomplete Data. GAO/RCED-00- 54. Washington, DC: GAO. Gibbons, R. D. In Press. An alternative statistical approach for performing water quality impairment assessments under the TMDL program. Gibson, G. R., M. L. Bowman, J. Gerritsen, and B. D. Snyder. 2000. Estuarine and Coastal Marine Waters: Bioassessment and Biocriteria Technical Guid- ance. EPA 822-B-00-024. Washington, DC: EPA Office of Water. Helsel, D. R., and R. M. Hirsch. 1992. Statistical Methods in Water Resources. Amsterdam: Elsevier. 522 p. Hirsch, R. M. 1982. A comparison of four record extension techniques. Water Resources Research 15:1781–1790. Hughes, R. M., and T. Oberdorff. 1999. Applications of IBI concepts and met- rics to waters outside the United States and Canada. Pages 79–93 in T. P. Simon, editor. Assessing the Sustainability and Biological Integrity of Water Resources Using Fish Communities. Boca Raton, FL: CRC Press. 

Waterbody Assessment: Listing and Delisting 65 ITFM (Intergovernmental Task Force on Monitoring Water Quality). 1992. Ambient water quality monitoring in the United States: first year review, evaluation, and recommendations. Washington, DC: Interagency Advisory Committee on Water Data. ITFM. 1993. Ambient water quality monitoring in the United States: second year review, evaluation, and recommendations. Washington, DC: Inter- agency Advisory Committee on Water Data. ITFM. 1995. The strategy for improving water-quality monitoring in the United States. Final report of the Intergovernmental Task Force on Moni- toring Water Quality. Washington, DC: Interagency Advisory Committee on Water Data. Jackson, L. E., J. C. Kurtz, and W. S. Fisher, editors. 2000. Evaluation Guide- lines for Ecological Indicators. EPA/620/R-99/005. Research Triangle Park, NC: EPA Office of Research and Development. Jennings, M. J., L. S. Fore, and J. R. Karr. 1995. Biological monitoring of fish assemblages in Tennessee Valley Reservoirs. Regulated Rivers: Research and Management 11:263–274. Karr, J. R. 1981. Assessment of biotic integrity using fish communities. Fish- eries 6(6):21–27. Karr, J. R. 1991. Biological Integrity: A Long-Neglected Aspect of Water Resource Management. Ecological Applications 1:66–84. Karr, J. R. 1998. Rivers as sentinels: Using the biology of rivers to guide land- scape management. Pages 502–528 in R. J. Naiman and R. E. Bilby, eds. River Ecology and Management: Lessons from the Pacific Coastal Ecosys- tems. New York: Springer. Karr, J. R. 2000. Health, integrity, and biological assessment: The importance of whole things. Pages 209–226 in D. Pimentel, L. Westra, and R. F. Noss, editors. Ecological Integrity: Integrating Environment, Conservation, and Health. Washington, DC: Island Press. Pgs. 214–215. Karr, J. R., and D. R. Dudley. 1981. Ecological perspective on water quality goals. Environmental Management 5:55–68. Karr, J. R., and E. W. Chu. 1999. Restoring Life in Running Waters: Better Biological Monitoring. Washington, DC: Island Press. Karr, J. R., and E. W. Chu. 2000. Sustaining living rivers. Hydrobiologia 422/423:1–14. Karr, J. R., K. D. Fausch, P. L. Angermeier, P. R. Yant, and I. J. Schlosser. 1986. Assessment of biological integrity in running waters: A method and its rationale. Illinois Natural History Survey Special Publication 5. Illinois Natural History Survey, Urbana, Illinois. McDonough, T. A., and G. D. Hickman. 1999. Reservoir fish assemblage in- dex development: A tool for assessing ecological health in Tennessee Val- ley Authority impoundments. Pages 523–540 in T. P. Simon, editor. As- sessing the Sustainability and Biological Integrity of Water Resources Us- ing Fish Communities. Boca Raton, FL: CRC Press. 

66 Assessing the TMDL Approach to Water Quality Management National Research Council (NRC). 2000. Watershed Management for Potable Water Supply: Assessing the New York City Strategy. Washington, DC: National Academy Press. Novotny, V. 1999. Integrating diffuse/nonpoint pollution control and water body restoration into watershed management. Journal AWRA 35(4):717–727. Novotny, V., and H. Olem. 1994. Water Quality: Prevention, Identification and Management of Diffuse Pollution. New York: Van Nostrand - Reinhold (distributed by Wiley). Ohio EPA (Environmental Protection Agency). 1988. Biological Criteria for the Protection of Aquatic Life, volumes 1–3. Columbus, OH: Ohio EPA Ecological Assessment Section, Division of Water Quality Monitoring and Assessment. Ohio EPA. 1990. Ohio water resource inventory, volume I, summary, status, and trends. Rankin, E. T., Yoder, C. O., and Mishne, D. A. (eds.). Colum- bus, OH: Ohio EPA Division of Water Quality Planning and Assessment Ohio EPA. 1999. Total maximum daily load TMDL team report. Columbus, OH: Ohio EPA Division of Surface Water. 139 pp. Ohio EPA. 2000. Ohio EPA five-year surface water monitoring strategy: 2000–2004 (draft). Columbus, OH: Ohio EPA Division of Surface Water, Ecological Assessment Unit. Plafkin, J. L., M. T. Barbour, K. D. Porter, S. K. Gross and R. M. Hughes. 1989. Rapid Bioassessment Protocol for Use in Stream and Rivers: Benthic Macroinvertebrates and Fish. EPA 440/4-89/001. Washington, DC: EPA. Rankin, E. T., and C. O. Yoder. 1990. A comparison of aquatic life use im- pairment detection and its causes between an integrated, biosurvey-based environmental assessment and its water column chemistry subcomponent. Appendix I, Ohio Water Resource Inventory (Volume 1). Columbus, OH: Ohio EPA, Division of Water Quality Planning Assessment. 29 pp. Singh, K. P., and G. S. Ramamurthy. 1991. Harmonic Mean Flows for Illinois Streams. Champaign, IL: Illinois State Water Survey. Smith, R. A., G. E. Schwarz, and R. B. Alexander. 1997. Regional interpreta- tion of water-quality monitoring data. Water Resources Research 33(12):2781–2798. Smith, E. P., K. Ye, C. Hughes, and L. Shabman. 2001. Statistical assessment of violations of water quality standards under Section 303(d) of the Clean Water Act. ES&T 35:606–612. Tampa Bay National Estuary Program. 1996. Charting the Course—The Com- prehensive Conservation and Management Plan for Tampa Bay. Wright, J. F., D. Moss, R. T. Clarke, and M. T. Furse. 1997. Biological as- sessment of river quality using the new version of RIVPACS (RIVPACS III). Pages 102–108 in P. J. Boon and D. L. Howell (eds). Freshwater Qual- ity: Defining the Indefinable? Scottish Natural Heritage, Edinburgh.Norris, R. H., B. T. Hart, M. Finlayson, and K. R. Norris (eds). 

Waterbody Assessment: Listing and Delisting 67 Wright, J. F., P. D. Armitage, and M. T. Furse. 1989. Prediction of invertebrate communities using stream measurements. Regulated Rivers: Research and Management 4:147–155. Yoder, C. O. 1997. Important elements and concepts of an adequate state wa- tershed monitoring and assessment program. ASIWPCA Standards & Monitoring, Ohio EPA Tech. Bull. MAS/1997-7-1. Columbus, OH: Ohio EPA Division of Surface Water. Yoder, C. O., and E. T. Rankin. 1995. Biological criteria program development and implementation in Ohio, pp. 109–144, in W. Davis and T. Simon (eds.). Biological Assessment and Criteria: Tools for Water Resource Planning and Decision Making. Boca Raton, FL: Lewis Publishers. Yoder, C. O., and E. T. Rankin. 1998. Biological response signatures and the area of degradation value: new tools for interpreting multimetric data, pp. 263–286. in W. Davis and T. Simon (eds.). Biological Assessment and Criteria: Tools for Water Resource Planning and Decision Making. Boca Raton, FL: Lewis Publishers.