Water: Total Maximum Daily Loads (303d)
Appendix D - Selected Technical Considerations
The TMDL Process
When developing a TMDL, design conditions are those critical conditions that must be specified in order to determine attainment of water quality standards. In specifying conditions in the waterbody, an attempt is made to use a reasonable "worst case" condition. For example, stream analysis often uses a low flow (e.g., 7-day low flow, once in 10-years commonly known as 7Q10 or biologically-based 4-day 3-year flows) high temperature design condition.
In situations where nonpoint source loadings at wet weather flow conditions are more significant than the point source loadings, the use of low flow-related design conditions is inappropriate. Wet weather flow conditions may be appropriate for analysis of nonpoint and intermittent point source discharges such as storm sewers. Other factors such as rainfall intensity and duration, time since previous rainfall, pollutant accumulation rates, and stream flow previous to rainfall should be considered in selecting design conditions for nonpoint source analysis. In some instances (e.g., carcinogenic pollutants), it is appropriate to use the harmonic mean flow to estimate loading capacity.
Often conditions of best management practices may be specified for factors other than physical conditions. For example, assumptions about cropping patterns, logging rates, or grazing practices may be necessary to determine the pollution loading estimates of a waterbody. Design conditions are less standardized for these factors and a reasonable worst case condition often must be developed on a case-by-case basis.
In general, for point sources, continuous discharges present the greatest stress under low flow, dry weather conditions. For pollutants transported in runoff, critical conditions will be rainfall-related, but may occur under a variety of flow conditions. For NPSs or intermittent point sources, generally, high flow, wet weather conditions need to be evaluated. For carcinogenic pollutants, harmonic mean flows may be appropriate. Additional details for selecting design conditions are provided in technical guidance.33
When the analyst is calculating a numerical TMDL, several mathematical models can be used to evaluate alternative pollutant loading scenarios. Models supported by the EPA Center for Exposure and Assessment Modeling (CEAM) are summarized in Appendix E. While it is beyond the scope of this guidance to provide a detailed rationale for model selection, the following briefly presents a discussion on model characteristics and selection.
Models can be characterized in numerous ways such as by their data requirements, ease of application, etc. This section summarizes models based on four categories: temporal characteristics, spatial characteristics, specific constituents and process simulated, and transport processes.
- Temporal characteristics - This includes whether the model is steady-state (inputs and outputs constant over time), time-averaged (for example, tidally-averaged), or dynamic. If the model is dynamic, an appropriate time step needs to be selected. For example, streams may require short time steps (hourly or less) while lakes, which typically have residence times in excess of weeks, can generally be modeled with longer time steps (e.g., daily or more). Similarly, loads from NPS models are often lumped together into event or annual loadings.
- Spatial characteristics - This includes the number of dimensions simulated and the degree of spatial resolution. In most stream models, one-dimensional models are used since typically vertical and horizontal gradients are small. For large lakes and estuaries, two- or three-dimensional models may be more appropriate because both vertical and horizontal concentration gradients commonly occur. Segmented or multiple catchment models may be more appropriate for heterogeneous watersheds, whereas, lumped single-catchment models are more appropriate for homogeneous or less complex situations.
- Specific constituents and processes simulated - Models vary in the types of constituents and processes simulated and in the complexity of the formulations used to represent each process. For example, simple DO models include only reaeration and BOD decay while more complex models include other processes such as nitrification, photosynthesis, and algal respiration.
- Transport processes - These include advection, dispersion, runoff, interflow, ground water interactions, and the effects of stratification on these processes. Most river models are concerned only with downstream advection and dispersion. Lake and estuary models may include advection and dispersion in one or more dimensions, as well as the effects of density stratification. For toxic modeling, it may be important to use models which account for near-field mixing since many of these pollutants may exert maximum toxicity close to the point of discharge. To incorporate both point and nonpoint sources into TMDLs, it will be important to consider integrated watershed models.
A model should be selected based on its adequacy for the intended use, for the specific waterbody, and for the critical conditions occurring at that waterbody. While the selection of an appropriate model should be made by a water quality analyst, it is useful for program managers to be familiar with the decisions which must be made. Four basic steps have been identified that an analyst would go through to select an appropriate model:
- Identify models applicable to the situation.
- Define the appropriate level of analysis.
- Incorporate practical constraints into the selection criteria.
- Select a specific model.
Identify models applicable to the situation. An obvious choice for narrowing the selection of an appropriate model is based on the waterbody type (river, estuary, or lake) and the type of analysis (BOD/DO, toxics, etc.) A preliminary list of models may also be screened by selecting models which consider the appropriate constituents and processes that are important for the pollutant being studied.
Define the appropriate type of analysis. Four types of models are:
- Simple calculator models - These include dilution and mass balance calculations, Streeter-Phelps equations and modifications thereof, analytical solutions to transport equations, steady-state nutrient loading models, regression models, and other simplified modeling procedures that can be performed on desk top calculators.
- Steady state computer models - These models compute average spatial profiles of constituents along a river or estuary assuming everything remains constant with time, including loadings, upstream water quality conditions, stream flow rates, meteorological conditions, etc.
- Quasi-dynamic models - These models are a compromise between steady-state models and dynamic models. Quasi-dynamic models assume most of the above factors remain constant, but allow one or more of them to vary with time, for example waste loading rates or stream flow rates. Some of the models hold the waste loading and flow rates constant, but predict effects such as the diurnal variations in dissolved oxygen due to algal photosynthesis and respiration.
- Dynamic models - These models predict temporal and spatial variations in water quality due to varied loadings, flow conditions, meteorological conditions, and internal processes within the watershed or waterbody. Dynamic models are useful for analyzing transient events (e.g., storms and long term seasonal cycles) such as those important in lake eutrophication analyses.
The above model types are listed in order of increasing complexity, data requirements, and cost of application. In addition, lognormal probabilistic models and Monte Carlo simulation techniques have been used to modify some of the above approaches. Probabilistic models use lognormal probability distributions of model inputs to calculate probability distributions of model output. Since this method does not incorporate fate and transport processes, it can only be used to predict the concentration of a substance after complete mixing and before decay or transformation significantly alters the concentration. Monte Carlo simulations combine probabilistic inputs with deterministic models. A fate and transport model is run a large number of times based on randomly selected input values. The output from these models are then rank ordered to produce a frequency distribution. These frequency distributions may then be compared to instream criteria (e.g., criteria maximum concentration (CMC) and criteria continuous concentration (CCC)) to determine if water quality standards are met.
Incorporate practical constraints. In general, the analyst should consider the data requirements for each level of analysis, the availability of historical data, the modeling effort required for each level of analysis, and available resources. Availability of historical data for calibration and verification is one of the key cost savings considerations.
Select a specific model. The analyst should consider model familiarity, technical support and model availability, documentation quality, application ease, and professional recognition and acceptance of a model.
Pollutant Allocation Schemes
Individual States use various load allocation schemes appropriate to their needs and may specify that a particular method be used. Methods of allocating loads have been historically applied to point sources. Application of these methodologies to nonpoint sources has not been well studied to date. Three common methods for allocating loads (equal percent removal, equal effluent concentrations, and a hybrid method) are discussed below. Other methods are detailed in another EPA document.34 The first method is equal percent removal and exists in two forms. In one, the overall removal efficiencies of the sources are set so they are all equal. In the latter, the incremental removal efficiencies beyond the current discharge are equal. This method is appropriate when the incremental removal efficiencies are relatively small, so that the necessary improvement in water quality can be obtained by minor improvement in treatment at each point source, at little cost.
The second common allocation method specifies equal effluent concentrations. This is similar to equal percent removal if influent concentrations at all sources are approximately the same. However, if one source has substantially higher influent levels, then equal effluent concentrations will require higher overall treatment levels than the equal percent removal approach.
The third commonly used method of allocating loads can be termed a hybrid method. With this method, the criteria for waste reduction may not be the same from one source to the next. One source may be allowed to operate unchanged while another may be required to provide the entire load reduction. More generally, a proportionality rule may be assigned that requires the percent removal to be proportional to the input source loading or flow rate.
TMDLs are particularly critical for waterbodies when the effect from multiple pollution sources overlap. The key concern associated with multiple point or nonpoint pollution sources is the potential for combined impacts. To perform this analysis, it may be necessary to apply near-field mixing models (mixing zone analysis) in addition to a far-field model which considers pollutants from numerous point or nonpoint sources (after the mixing zone). A recommended procedure for evaluating toxicity from multiple discharges is summarized in EPA guidance.35
Where appropriate and technically feasible, certain cost-effective benefits may be gained by making tradeoffs among wasteload allocations. Such a practice is similar to what would be done during the initial considerations of tradeoffs of loads between point and nonpoint sources. In the case of watershed or estuary management, this may be particularly useful to achieve pollution reduction in the most cost-effective manner possible.
The incentive for trading load allocations is to achieve the required level of control by choosing to control one pollutant source over another. Technological feasibility, economic issues, and regulatory authority are all factors to consider when trading allocations. For example, to reduce nutrient loads to a receiving water, nonpoint source controls that can be adequately maintained and enforced, may be much more cost effective than increasing the level of control on a point source discharger.
Pollutant trades are most likely to occur between point and nonpoint sources. However, where effluents from different point source dischargers are comparable, trades may be acceptable so long as water quality standards (including antidegradation regulations and policies) and minimum applicable technology-based controls are met. Similarly, tradeoffs between nonpoint sources are also acceptable.
The Dillon Reservoir (west of Denver, Colorado) is an example of point and nonpoint source phosphorus load tradeoffs. In this example, the cost associated with point source reduction was $1.5 million per year, whereas the cost associated with NPS controls was $0.2 to $1.0 million per year. Because of this cost differential, tradeoffs allowed publicly-owned treatment works to achieve reductions in phosphorus loads to the Dillon Reservoir by controlling NPSs rather than expanding the sewage treatment system.
Persistent and/or Highly Bioaccumulative Toxic Pollutants
Persistent and/or bioaccumulative toxic pollutants require special attention during analysis of toxicity and TMDL development. The primary concern is that toxic pollutants that enter a waterbody at levels that are non-toxic in the water column may accumulate in sediment or aquatic life. These pollutants may then adversely affect aquatic/wildlife or pose a risk to humans by exposure to hazardous chemicals through consumption of contaminated fish or shellfish. Chemicals that bioaccumulate at high rates include some metals, organic compounds, and organometallic compounds. Current technical guidance for wasteload allocation (see Appendix A) summarize a number of models which are appropriate for modeling the fate and transport of toxics in streams/rivers, lakes, and estuaries. Additional details for assessing and controlling risk have been addressed in technical support documentation.
Use of Two-number Criteria
Because of inherent variation in effluent and receiving water flows and pollutant concentrations, specifying a concentration that must not be exceeded at any time or place may not be appropriate for the protection of aquatic life. The format usually selected for expressing water quality criteria to protect aquatic life consists of recommendations concerning concentration magnitudes, duration of averaging periods, and average frequencies of allowed excursions. Use of this magnitude-duration-frequency format allows water quality criteria for aquatic life to be adequately protective without being as overprotective as if criteria were expressed using a simpler format. In many cases, these considerations are evaluated during the standards setting process and TMDLs are used to develop controls that result in attainment of applicable water quality standards.
Duration of exposure considers the amount of time organisms will be exposed to toxicants. It is expressed as that period of time over which the instream concentration is averaged for comparison with criteria concentrations. Frequency is defined as how often exposures that exceed the criteria can occur during a given period of time (e.g., once every three years) without unacceptably affecting the community. To account for acute toxic effects, States may adopt acute criteria expressed as the criteria maximum concentration (CMC) occurring in a one-hour averaging period. Similarly, chronic criteria expressed as the criteria continuous concentration (CCC) should be developed as toxicant concentrations which should not be exceeded over longer periods of time. For the purposes of modeling, the ambient concentration should not exceed the CMC more than once every three years. (If the biological community is under stress because of spills, multiple dischargers, or has a low recovery potential, or if a local species is very important, the frequency should be decreased.)
Although these criteria are mostly used for application to low flow conditions, the toxicological basis for the criteria is equally valid for high flow conditions. It is important for States to protect designated water uses during all flow conditions; therefore, the two-number criteria should be used for all flow conditions unless separate guidance for adopting wet weather criteria is available. However, States should apply duration and frequency parameters to account for the high flow, intermittent nature of nonpoint source loadings.
The problems associated with clean and contaminated sediment are not the same. Clean sediment can impair fish reproduction by silting-up spawning areas, and can increase turbidity. Draft (clean) sediment criteria have been developed in Idaho that include turbidity, inter-gravel dissolved oxygen, and cobble embeddedness. The criteria developed may be most appropriate for salmonid streams, but the framework may have wide application. The major concerns regarding contaminated sediment are pollutant releases to the water column, bioaccumulation, and biomagnification. Sediment criteria being developed by EPA have centered on evaluating and developing an understanding of the principal factors that influence the sediment/contaminant interactions with the water column (Equilibrium Partitioning Approach). (The Science Advisory Board will be reviewing methods for establishing sediment criteria for metal contaminants and procedures for establishing standardized bioassays in 1991.) Through such an understanding, exposure estimates of benthic and other organisms can be made. Chronic water quality criteria, or possibly other toxicological endpoints, can then be used to predict potential biological effects.
In some cases, sediment criteria alone would be sufficient to identify and to establish clean up levels for contaminated sediments. In other cases, the sediment criteria should be supplemented with biological or other types of analysis before clean-up decisions can be made. Additionally, ground water inputs through sediments should be distinguished from inputs from the sediment alone, so that proper control measures are implemented.
33.USEPA. 1985. Technical Support Document for Water Quality-based Toxics Control. OW/OWEP and OWRS, EPA 440/4-85-032. Washington, D.C. A revised draft (April 23, 1990) is available and will replace the 1985 Guidance when finalized. Back.
34.USEPA. 1985. Technical Support Document for Water Quality-based Toxics Control. OW/OWEP and OWRS, EPA 440/4-85-032. Washington, D.C. A revised draft (April 23, 1990) is available and will replace the 1985 Guidance when finalized. Back.
35.USEPA. 1985. Techical Support Document for Water Quality-based Toxics Control. OW/OWEP and OWRS, EPA 440/4-85-032. Washington, D.C. A revised draft (April 23, 1990) is available and will replace the 1985 Guidance when finalized. Back.