Front matter for Total Maximum Daily Load Development and Implementation

A total maximum daily load (TMDL) is the maximum allowable loading limit of a particular water quality constituent such as nutrients, sediment, temperature, salinity, pathogen counts, and others that can enter an impaired water system (receiving waterbody) identified under Section 303(d) of the Clean Water Act (CWA) of 1972, so that the applicable water quality standards can be met. In a regulatory context, the TMDL facilitates a plan of action to manage both point source (PS) and non-point-source (NPS) pollution. The minimum elements in a TMDL required by the USEPA's Guidelines for Reviewing TMDLs under Existing Regulations issued in 1992 are the following:

As is the case with most natural systems, there are no experimental or direct methods to derive a TMDL. Developing a TMDL is a systematic process that quantifies the extent of pollutant loading to a waterbody and develops a plan to address the water quality impairment for which the use of watershed and receiving water quality models is key. As documented in the USEPA Assessment and TMDL Tracking and Implementation System (ATTAINS), 76,127 TMDLs were developed across the United States during the period 1975 through 2017 employing a wide variety of analytical or empirical procedures/methods, simple models, watershed models, and receiving water quality models. The tools were chosen by the individual TMDL developers without the benefit of clear guidance or standardized protocols that might have guided the analysis procedures and led to better choice of model.

Although guidance manuals have been developed by USEPA for the purpose of pollutant load allocations, neither USEPA nor any state agency has developed specific guidance on TMDL analysis and modeling. The EWRI Watershed Management Technical Committee of the ASCE first chartered the Total Maximum Daily Load Analysis and Modeling Task Committee (TMDL-TC) in 2011 to explore the development of an ASCE Manual of Practice (MOP) on TMDL analysis and modeling for the development and implementation of TMDLs. After a careful review of current and past practices, the TMDL-TC confirmed the inadequacy of existing guidance for TMDL analysis and modeling. As stated in an ASCE book authored by the TMDL-TC in 2017, an ASCE MOP was both timely and critical to achieve the water quality objectives established by the CWA. In 2019, as an interim step, the TMDL-TC members published a special collection of papers in the ASCE Journal of Hydrologic Engineering that sought to define the scientific basis more clearly for TMDL modeling and describe the state-of-the-art. The MOP that followed the 2017 ASCE book and the special collection of associated technical papers relates the state-of-the-practice to the state of the art in TMDL analysis and modeling. It was not always possible to draw a clear distinction between these designations, given the rapid innovation in the embrace of technologies such as GIS and remote sensing and the realization that certain current practices may become obsolete in the not-too-distant future. The MOP addresses this and other relevant topics/subjects in each of the MOP chapters.

The 12 chapters contained in the MOP are on TMDL analysis and modeling and other related topics needed in the development and implementation of TMDL arranged in an order corresponding to a typical TMDL development workflow to guide the readers. The chapters describe the various watershed and receiving water models; discuss the integration or linking of these models for complex waterbody systems; simple analytical and empirical procedures/methods and models; critical condition determinations; data types and sources; calibration, validation, and performance quantifications of the models; model uncertainty analysis and margin of safety quantification for TMDL developments; a database of USEPA-approved TMDL reports and a computer search tool to search those reports; a model selection protocol for TMDLs, the principal steps involved in modeling for TMDL development and implementation planning along with an implementation planning case study; and current analysis and modeling practices for TMDL implementation planning.

Brief summaries of the individual chapters are given below:

Chapter 1 Introduction: The chapter provides a brief history of the TMDL program and the overall TMDL process. It expands on the TMDL definition and rationale and purpose of the MOP that is summarized in the section above.

Chapter 2 Watershed Models: The chapter presents watershed models that are primarily needed and used in TMDL development and implementation. Fourteen prevalent watershed models, listed subsequently, were selected based on their documentation including user manuals, successful applications on various watersheds, evaluations, and acceptances found in the literature. In Chapter 2, these 14 watershed models are further evaluated to determine their capabilities suitable for TMDL development and implementation.

Brief descriptions of these watershed models, including sources, capabilities, and applicability, are presented in Chapter 2. General information such as intended watershed (agriculture, urban or mixed), simulation type (event, continuous or both), simulated variables, uncertainty analysis, graphical user interface, and availability of the 14 models is presented in a comprehensive tabular form to compare the attributes of the models. The theoretical basis for the various hydrologic and water quality processes (phases or components) for each of the 14 models are tabulated for easy comparison of the models. Looking at the theoretical basis of models is important as it suggests expected performance and accuracy and dictate model features such as structure, input data, and model parameters. Among these, the water routing procedures derived from approximations of the St. Venant shallow water equations are the primary drivers of the models, based on which the models are compared and ranked on pictorial graphs based on the level of physical basis or accuracy against complexity. For example, the GSSHA and MIKE SHE models rank high on the simulation of overland flows and the SWMM ranks high in channel or pipe flow routings using the most accurate physical process–based equations—also being the most numerically complex. The GWLF model ranks at the bottom with respect to both flow simulation methodologies. The remainder of the models falls in between the two end members, although the DWSM model is the most computationally efficient among the kinematic wave models. Notable strengths and limitations of the watershed models for TMDL development and implementation are tabulated. This valuable compilation of information on the subject models is unique and should aid comparisons of relative model reliability and help guide selections for TMDL and other environmental applications.

Of the 14 watershed models presented in Chapter 2, applications of 12 (HSPF, LSPC, GWLF, SWAT, SWMM, WARMF, AGNPS, ANSWERS, HEC–HMS, KINEROS, WAM, and MIKE SHE) can be found in TMDL studies. However, caution should be exercised to their appropriateness for all applications - misuses of models are sometimes found in past TMDL study reports. Although the DWSM and GSSHA models have not been used in TMDL applications, these two models have been evaluated and used in other watershed management and best management practices (BMP) evaluation studies and, therefore, would be suitable for future TMDL development and implementation studies.

Chapter 2 outlines future research on watershed modeling. This research should include further comparisons of models performance based on other key aspects such as simulation capabilities of processes, uncertainty analyses, required resources, and ability to characterize pollutant loading from complex watersheds. Robust physically based algorithms, uncertainty analysis capabilities, and improvements in the facility of use of remotely sensed and high-resolution data for model development are recommended to be part of the future advancement in modeling technology, application, and solution rigor.

Chapter 3 Receiving Water Quality Models: The chapter describes the following 11 receiving water quality models that were previously used in TMDL development and implementation planning:

The sources, simulation capabilities, and applications of each of the 11 receiving water quality models are presented and briefly described in Chapter 3. Availability of these models in the public domain and specific capabilities of each of the models, such as 1D, 2D, or 3D hydrodynamic routings; types of modeling systems (rivers, lakes, reservoirs, or estuaries); and simulating pollutants (temperature, salinity, dissolved oxygen, carbon, nitrogen, phosphorous, algae, sediment, pathogen, toxics, and metals) are summarized in a table/matrix facilitating comparisons and the best selections of the models for applications in various TMDL studies (analysis). Like watershed modeling, receiving water quality modeling has also been researched and developed for over half a century. Thus, in this instance, the state-of-the-practice has caught up with the state-of-the-art. Chapter 3 serves as a valuable resource to understand the capabilities of the 11 models and match those with the conditions of the impairments and the impaired waterbodies for selecting the best model(s) for TMDL development and implementation.

Chapter 4 Integrated Modeling Systems and Linked Models: In this chapter, integrated modeling systems and linked models are described with examples. Linking of different models is needed to represent complex natural systems such as watersheds and their estuaries when such systems cannot be represented by a single model. Linkages can be static or dynamic. A static linkage takes output from one model and uses it as input to a second model. A dynamic linkage can be bidirectional, where information from each time step transfers back and forth between the models and affects simulations of all. Integrated modeling systems and linked models can satisfy multiple model application criteria because they include the capabilities of multiple models, provide various software tools, and include data and analysis support. Chapter 4 presents two widely used integrated modeling systems, briefly summarized in the following two paragraphs, along with a listing or brief descriptions of example applications in TMDL studies. Also, as summarized subsequently, three example applications of linked models are presented in Chapter 4 along with an important discussion on the performance evaluations of integrated modeling systems and linked models in TMDL applications.

In addition to citing a few applications of linked models, Chapter 4 describes the following three example applications of linked models in TMDL development and implementation:

Chapter 5 Simple Models and Methods: The chapter presents simple models and methods that are used to support assessment of the relative significance of different pollutant sources and to guide decisions for management plans. These models and methods are typically derived from empirical relationships between the physiographic characteristics of the watershed and pollutant exports and are often used when data limitations and budget and time constraints preclude using more detailed approaches. These approaches fit the needs of stakeholders tasked with implementing TMDLs where the focus is developing an understanding of the sources and quantification of pollutants causing impairments in a waterbody and determining how best to reduce the loads from prioritized pollutant sources that had been identified in the TMDL from previous monitoring, modeling, and analysis. The following is a list of 13 simple models and methods available to simulate watersheds (10 models/methods) and receiving waterbodies (three models/methods) that are discussed in Chapter 5:

The aforementioned procedures and models are limited in many ways and, therefore, a knowledge of these limitations is critical in the successful application of the procedures and the models. One major advantage is that these tools can provide a rapid means of identifying critical loading areas with minimal effort and data requirements. A major disadvantage is that only gross estimates of nutrient loads can be provided and are of limited value for determining loads on a seasonal or finer timescale. Another disadvantage is that these techniques may be too generalized and of limited use for evaluating the effectiveness of various BMPs. However, a strategy that starts with simple models in the initial phases and then progresses to more complex frameworks as additional data are collected, and as appropriate remedial measures are assessed, can be successful and cost-effective. Thus, simple models and methods can play an important role in TMDL development and evaluation as model selection invariably involves trade-offs involving model complexity, required reliability, cost, and time.

Chapter 6 Critical Condition Determination for Total Maximum Daily Load Modeling: In this chapter, critical condition determination in developing a TMDL and the appropriate model or methodologies are discussed. According to the Clean Water Act, the USEPA TMDL program requires a consideration of the critical conditions for streamflow, loading, and water quality parameters for the development of TMDLs. The intent is to ensure that the water quality of an impaired waterbody is protected during times when it is most vulnerable, such as during low- or high-flow periods. The following four methods have been used in critical condition modeling and analysis in TMDL development and implementation (examples are provided in this chapter):

Low-flow analyses are performed using steady-state models. For TMDLs related to impaired waterbodies affected by pollutants from both point and nonpoint sources, a dynamic continuous model is commonly used along with the selection of a representative period to account for the varying climatic and hydrologic conditions in a watershed (e.g., to cover dry, wet, and average years of precipitation). The load-duration curve method is a simple screening tool for problem characterization as part of TMDL development. The event-based critical flow-storm method is essentially a risk-based approach that explicitly addresses critical condition as a combination of streamflow, the magnitude of a storm event, and initial watershed conditions. Decision-making authorities could substantially benefit from information on risk and the associated probabilities of various combinations of rainfall, river flow conditions, and the pollutant loads generated under these combinations.

Chapter 7 Model Data, Geographic Information Systems, and Remote Sensing: The chapter outlines the data requirements in modeling for TMDL development and describes the geographic information system (GIS) and remote sensing (RS) technologies available for the acquisition of high-resolution spatial and temporal data needed to set up, initialize, and run detailed watershed models. Data requirements for a TMDL model setup are tabulated; also tabulated are data requirements for running the TMDL models. An overview of the GIS and an up-to-date evolution of the RS are presented. Also discussed are the specific data needs for certain TMDL applications and how the RS platforms can be matched to these specific data requirements. RS has added another dimension to measuring and retrieving hydrologic and atmospheric data for water quality and hydrologic modeling. Along with the advancements in RS technology, acquisition, and processing of large amounts of information are being managed using robust and easy-to-use GIS software both in the commercial and in the public domain. ArcGIS and other GISs are ubiquitous and have evolved to a point where these tools are extensively used in watershed and impaired waterbody modeling. Chapter 7 provides an overview of the more common GIS and RS technologies used in hydrologic, water quality, and TMDL modeling applications.

Chapter 8 Model Calibration and Validation: The chapter presents and discusses state-of-the-art approaches in model calibration and validation as found in the literature as a guide for the calibration and validation of models applied in TMDL development and implementation. The chapter focuses on watershed models, although the same are applicable to receiving water models as well. Model data requirements and sources of the model data presented in Chapter 7 are further expanded and refined in this chapter by tabulating various types of data and sources and providing website links of the well-maintained data sources. Discussions on data management and guidance on pre-calibration; calibration, including manual calibration and auto-calibration; and validation are presented. The principal parameters requiring calibration in the 14 watershed models presented in Chapter 2 (a few exceptions) are tabulated separately for hydrologic and water quality simulations, which are valuable resources to modelers. Measuring model performance or calibration success is discussed, including a tabulation of various formulas found in the literature for the goodness-of-fit measures of watershed and receiving water quality model predictions. Acceptance criteria of some of the measuring formulas or methods are also tabulated.

Chapter 9 Model Uncertainty Analysis and Margin of Safety: The chapter presents and discusses uncertainties in modeling, and accounting these in a Clean Water Act mandated margin of safety (MOS) in TMDL development and implementation, along with several estimation methods, including explicit, implicit, and risk-based MOS methods. In addition, a survey of sampled 17 TMDL reports is presented in a table compiling the TMDL case study, the state in which they are located, impairment (pollutants), the TMDL model or method used, and the MOS estimation method used. A comprehensive review of uncertainty analysis methods, including MOS estimations, are presented and discussed, and practitioners may find them helpful in TMDL development and implementation. Uncertainty analysis methods applied in 19 water quality modeling studies are tabulated listing the methods and the models used, the impairments (pollutants), and study references, a valuable resource for modelers or practitioners. Detailed descriptions and some of the equations of some of those methods are presented. The methods are (1) first-order variance analysis, (2) Monte Carlo method, (3) Bayesian Monte Carlo analysis, (4) the generalized likelihood uncertainty estimation (GLUE) method, (5) the Markov Chain Monte Carlo method, and (6) Kalman Filtering.

Chapter 10 USEPA Total Maximum Daily Load Report Archive and Report Search Tool: The chapter briefly describes the USEPA TMDL report archive database and online system, called ATTAINS (Assessment and Total Maximum Daily Load Tracking and Implementation System), and presents a new online tool developed for searching TMDL reports for certain search categories such as specific pollutants (impairments), models, or others. Previously published TMDL reports may serve as helpful guides in developing new TMDLs. A web-based software (tool) called the “TMDL Report Selection” (TRS) tool was developed as part of the TMDL Analysis and Modeling Task Committee's efforts to search and analyze published TMDL reports found in ATTAINS. The TRS tool is a state-of-the-art search tool that promotes greater use of information found in reports of USEPA-approved TMDL projects in the USEPA archive ATTAINS.

In Chapter 10, the TRS tool and its application examples provide statistics from TMDL reports published during the period 1975 to 2017. These statistics include (1) the number of reports published in each of the 55 states; (2) number of reports published during each of the years 1975 to 2017; (3) number of reports using each of the selected methods or models including watershed models, receiving water quality models, and simple models; (4) number of reports addressing each of the impairments; and (5) number of reports using each of the selected methods or models that address each of the selected impairments. Guidance on using the TRS tool is also provided.

Chapter 11 Model Selection and Applications for Total Maximum Daily Load Development: In this chapter, a TMDL model selection protocol is presented with descriptions of its basis and various considerations (technical, management, and stakeholder and expert engagements) in the development process and summarizing these in an easy-to-read flowchart. The chapter also describes the process of applying models in the development and implementation of TMDLs and briefly describes a case study for evaluating TMDL load reduction scenarios in the form of an appendix in this chapter. It first lists the types of models used to develop TMDLs and then presents a holistic approach and a protocol to select a practical and dependable model for TMDL development and implementation. In this protocol, rigorous hypothesis testing of the selected model is emphasized, particularly if the model has not already been rigorously peer reviewed. The holistic protocol covers all aspects of the model application, from conceptualization and stakeholder collaboration to implementation planning and outreach. These considerations are translated into technical and management selection criteria and synoptic data requirements. Model performance evaluation, calibration, validation, and uncertainty estimation are all model application steps that influence selection of the model as described in Chapter 11. It also includes the modeling procedures necessary to support TMDL determination, allocation, and implementation planning. Looking into these steps ensures that the selected TMDL model is appropriate and likely to be successful and approved by the USEPA, which also issues National Pollution Discharge Elimination System (NPDES) permits. Other federal, state, and local agencies and stakeholders have roles in applying best management practices and other controls and overseeing the collection of synoptic data for calibration, validation, and determining model reliability. Finally, a discussion on the current state of the TMDL modeling practices and potential future innovations (developments) is presented.

Chapter 12 Modeling for Total Maximum Daily Load Implementation: The chapter discusses state-of-the-practice on the uses of models in TMDL implementation planning. Ideally, the same model used in TMDL development can also be used for TMDL implementation and related assessment; however, this is not always the case. Communities, organizations, agencies, and other stakeholders implementing a TMDL typically seek the most reliable and importantly the most cost-effective methods to achieve the load reductions required. They invariably intend to compare different scenarios that provide trade-offs among measures such as higher load reduction, lower cost, lower maintenance, available land, or easier implementation. In making these comparisons, the use of an easily understood and reliable model to estimate the load reduction from different planning scenarios is an important element of decision-making for TMDL implementation.

Chapter 12 discusses the uses of the following few watershed-specific models:

Chapter 12 also discusses two spreadsheet analysis packages:

Critical comparative information (structure, capabilities, and treatment practices) of the aforementioned six models are tabulated for comparisons. Finally, a brief description of custom modeling using GIS geodatabases is presented with the example of the integrated modeling tool for scenario development for the District of Columbia. From an overall implementation perspective, Chapter 12 summarizes a number of ways by which TMDL implementation can be carried out with a better probability of success, including an adaptive management approach for complex TMDLs.

This manual describes the state-of-the-practice and provides guidance to practitioners in selecting analytical tools and models in the development of a TMDL and its implementation plan. It includes detailed descriptions of a variety of watershed and receiving water quality models that can be used, highlighting recent advances in TMDL analysis and modeling.

This manual comprising 12 chapters is expected to serve as a useful resource to engineers, practitioners, regulators, and other water resource and watershed managers charged with developing and implementing TMDLs. The manual will also serve as a useful resource to planners, environmental scientists, students, educators, the concerned citizens, watershed stakeholders, and the public.

All chapters are based on state-of-the-art reviews of the respective topics, except Chapters 10 and 11, which introduce a state-of-the-art TMDL report search tool in the ATTAINS database and a state-of-the-art TMDL model selection protocol. All chapters serve as useful resources or tools in developing and implementing reliable and successful TMDLs.

As more TMDL studies result in implementation, the use of models for management planning and alternate analysis is increasing; for example, explicit incorporation of BMPs and low-impact development (LID) in the watershed and water quality models has advanced in recent years. The advent of machine learning techniques and other statistical data science technologies can assist communities and watershed stakeholders in achieving TMDL pollutant load reduction goals in a more sustainable and cost-efficient manner. Thus, this Manual of Practice expects to become a living document and needs to be updated in the future to account for potential advances in the practice of TMDL analysis and modeling for the development and implementation of TMDLs.