Selecting Reliable Models for Total Maximum Daily Load Development: Holistic Protocol
Abstract
Introduction
Statea | Number of reports | Number of TMDLs | Source |
---|---|---|---|
Alabama (AL) | 3 | 7 | USEPA ATTAINS |
Alaska (AK) | 3 | 6 | ADEC (2020) |
Arizona (AZ) | 3 | 11 | USEPA ATTAINS |
Arkansas (AR) | 5 | 23 | ADEQ (2020) |
California (CA) | 9 | 335 | USEPA ATTAINS |
Colorado (CO) | 4 | 16 | USEPA ATTAINS |
Washington, DC (DC) | 2 | 41 | District of Columbia Department of Energy and Environment (2021) |
Florida (FL) | 10 | 28 | Florida Department of Environmental Protection (2020) |
Georgia (GA) | 40 | 251 | Georgia Environmental Protection Division (2020) |
Hawaii (HI) | 1 | 3 | Hawaii Department of Health (2020) |
Idaho (ID) | 11 | 80 | Idaho Department of Environmental Quality (2020) |
Illinois (IL) | 19 | 71 | Illinois Environmental Protection Agency (2020) |
Indiana (IN) | 5 | 285 | Indiana Department of Environmental Management (2020) |
Iowa (IA) | 11 | 61 | Iowa Department of Natural Resources (2020) |
Kansas (KS) | 9 | 104 | USEPA ATTAINS |
Kentucky (KY) | 4 | 45 | Kentucky Energy and Environment Cabinet (2020) |
Louisiana (LA) | 1 | 1 | USEPA ATTAINS |
Maine (ME) | 1 | 21 | USEPA Region 1 |
Massachusetts (MA) | 25 | 99 | USEPA Region 1 |
Michigan (MI) | 2 | 2,107 | USEPA ATTAINS |
Minnesota (MN) | 22 | 481 | USEPA ATTAINS |
Mississippi (MS) | 1 | 2 | USEPA ATTAINS |
Missouri (MO) | 4 | 5 | USEPA ATTAINS |
Montana (MT) | 4 | 64 | USEPA ATTAINS |
New Hampshire (NH) | 6 | 8 | New Hampshire Department of Environmental Services (2020) |
New Jersey (NJ) | 1 | 3 | New Jersey Department of Environmental Protection (2021) |
New Mexico (NM) | 5 | 27 | USEPA ATTAINS |
New York (NY) | 7 | 7 | New York Department of Environmental Conservation (2020) |
North Carolina (NC) | 3 | 10 | North Carolina Department of Environmental Quality (2020) |
North Dakota (ND) | 4 | 9 | USEPA ATTAINS |
Oklahoma (OK) | 3 | 14 | Oklahoma Water Quality Division (2020) |
Oregon (OR) | 4 | 7 | Oregon Department of Environmental Quality (2020) |
Rhode Island (RI) | 1 | 2 | Rhode Island Department of Environmental Management (2020) |
South Carolina (SC) | 5 | 26 | USEPA ATTAINS |
South Dakota (SD) | 2 | 77 | USEPA ATTAINS |
Tennessee (TN) | 18 | 402 | Tennessee Department of Environment and Conservation (2020) |
Texas (TX) | 11 | 135 | Texas Commission on Environmental Quality (2020) |
Utah (UT) | 12 | 23 | Utah Department of Environmental Quality (2020) |
Vermont (VT) | 2 | 2 | USEPA ATTAINS |
Virginia (VA) | 11 | 97 | Virginia Department of Environmental Quality (2020) |
Washington (WA) | 9 | 265 | Washington State Department of Ecology (2020) |
West Virginia (WV) | 8 | 236 | West Virginia Department of Environmental Protection (2020) |
Wisconsin (WI) | 3 | 7 | Wisconsin Department of Natural Resources (2020) |
Wyoming (WY) | 1 | 19 | Wyoming Department of Environmental Quality (2020) |
Total | 315 | 5,523 | — |
Source: Data from USEPA (2020a).
Key Steps in TMDL Determination, Allocation, and Implementation Planning
Groups Involved in Determining the TMDL
Modeling for TMDL Determination, Allocation, and Implementation Planning
Type of TMDL model | Number of TMDL reports | Pollutants | Model reference |
---|---|---|---|
Empirical models | |||
Regression equation | 24 | Sediment, turbidity, total suspended solids, temperature, dissolved oxygen, pH, total phosphorous, ammonia, phosphate, chloride, noxious aquatic plants, bacteria, fecal coliform, E. coli, boron, aluminum, copper, selenium, silver, cadmium, mercury, uranium, polychlorinated bisphenyls, simazine | USEPA (1991) |
Load-duration curve | 101 | Sediment, turbidity, total suspended solids, total dissolved solids, nutrients, nitrate, chloride, pathogens, Escherechia coli, Fecal coliforms, enterococci, total phosphorous, boron, manganese, iron, copper, selenium, zinc, lead, uranium, atrazine, terbufos | USEPA (2017) |
Simple analytical models | |||
CNET | 1 | Total phosphorous, total suspended solids, Escherechia coli | Walker (1989) |
FLUX32 | 1 | Phosphorous | Jensen and Freihoefer (2012) |
ENSR-LRM | 4 | Phosphorous | New Hampshire Department of Environmental Services (2013, 2020), and Connecticut Department of Environmental Protection and ENSR (2004) |
STEPL | 11 | Dissolved oxygen, pH, total phosphorous, sediment, nutrients, simazine | USEPA (2020d) |
USLE/RUSLE2 | 11 | Sediment | USDA (2016) |
Unnamed mass-balance model | 14 | Sediment, total dissolved solids, dissolved oxygen, pH, nutrients, chloride, sulfate, pathogens, total coliform, fecal coliform, Escherechia coli, enterococci, aluminum, cadmium, selenium, zinc, lead, trash | USEPA (2018c) |
Receiving water quality models | |||
Unnamed 1D hydrodynamic model | 1 | Selenium | Shoemaker et al. (2005) |
Aquatox | 1 | Total phosphorous | USEPA (2018a) |
CE-QUAL-W2 | 1 | Temperature, total phosphorous | Cole and Wells (2018) |
RMA2 | 1 | Temperature, dissolved oxygen | Aquaveo (2019) |
SSTEMP | 2 | Temperature | Bartholow (2000) |
RBM-10 | 2 | Temperature | Yearsley (2009) |
RMA11 | 2 | Temperature | RMA (2018) |
SWMM | 2 | Total suspended solids, total phosphorous, Escherechia coli, copper, zinc, lead | USEPA (2020e) |
Unnamed 3D hydrodynamic model | 2 | Dissolved oxygen, mercury | Shoemaker et al. (2005) |
Unnamed 1D steady-state model | 5 | Dissolved oxygen, biochemical oxygen demand, selenium | Shoemaker et al. (2005) |
EFDC | 8 | Dissolved oxygen, ammonia, chlorophyll a | Hamrick and Wu (1997) |
Qual 2k | 13 | Dissolved oxygen | Chapra et al. (2008) |
BATHTUB | 40 | pH | Walker (2006) |
Watershed and integrated models | |||
HydroWAMIT | 1 | Total suspended solids, dissolved oxygen, pH, total phosphorous | Cerucci and Jaligama (2008) |
SLAMM | 1 | Sediment, total suspended solids, phosphorous, fecal coliform | PV and Associates (2019) |
WinSLAMM | 1 | Total phosphorous | PV and Associates (2019) |
WASP | 2 | Total suspended solids, dissolved oxygen, pH, total phosphorous | USEPA (2019b) |
GWLF | 5 | Sediment, siltation | Haith et al. (1992) |
LSPC | 7 | Sediment, temperature, bacteria, chlorophyll a | USEPA (2016) |
MDAS | 9 | pH, Chloride, fecal coliform, aluminum, manganese, iron, selenium | Jian et al. (2002) |
SWAT | 11 | Sediment, total suspended solids, turbidity, nutrients, total phosphorous, Escherechia coli, fecal coliform | Arnold et al. (2012) |
HSPF | 31 | Sediment, total suspended solids, turbidity, dissolved oxygen, biochemical oxygen demand, nutrients, total nitrogen, total phosphorous, chlorophyll a, Escherechia coli, polychlorinated bisphenyls | Bicknell et al. (2001) |
Total | 315 | — | — |
Note: CNET = Reservoir eutrophication modeling worksheet; ENSR-LRM = ENSR-lake response model; STEPL = spreadsheet tool for estimating pollutant load; USLE/RUSLE2 = universal soil loss equation/revised universal soil loss equation 2; RMA2 = resource management associates one dimensional/two dimensional hydrodynamic model; SSTEMP = stream segment temperature model; RBM-10 = river basin model-10; RMA11 = resource management associates two/three dimensional finite element model for water quality simulation; SWMM = storm water management model; HydroWAMIT = hydrologic watershed model integration tool; WASP = water quality analysis program; GWLF = generalized watershed loading function; LSPC = loading simulation program in C++; MDAS = mining data analysis system; SWAT = soil and water assessment tool; and HSPF = hydrological simulation program-fortran.
Protocol for Model Selection
Search for Existing Model Selection Protocol
Proposed Model Selection Protocol
Fundamental Model Selection Principles
Key Factors in the Model Selection Process
Technical Criteria
Principal criteria | Examples | Positive impacts | Negative impacts |
---|---|---|---|
Impaired waterbody | Waterbody type, hydrological conditions | Most system conditions represented by models that capture typical, critical (dry years or dry riverbeds), and extreme conditions (such as floods and inundation) | • Critical and extreme events missed by models that only cover typical range of flows • Episodic compliance failures |
Pollutants | Pollutants, loads, levels of impairment | • Well-constructed models by thorough accounting of PS and NPS through stakeholder engagement, surveys, and careful evaluation of GIS data sources • TMDL with proper load-reduction allocations or assimilation limits by incorporating cause-and-effect relationships between loads and water quality | • Incorrect specification of loads and impacts of BMPs from models designed without adequate domain expertise about PS and NPS loads • Poorly constructed load-reduction allocations or assimilation limits, which can lead to unrealistic TMDLs • Impacts of BMPs on load reduction not carefully considered |
Data requirements | Reliability | Well-calibrated models with well-documented error propagation | • Difficulty in calibrating and confirming models due to unreliable data • Difficulty in error propagation analyses |
Resolution | High-resolution data-driven models | Low-resolution approximate models | |
Coverage | Synoptic data coverage leading to model calibration and confirmation at several spatial locations and reduction of local and global errors | • Poorly sampled regions causing potential global imbalances in the model • Incorrect load allocations | |
Process-resolution | High-frequency data collection at fronts and gradients, among other physical processes, leading to well-resolved models with accurate representation of mechanistic processes | • Poorly resolved processes • Unaccounted sources | |
Access | Accurate data on NPS loads to build detailed TMDL models | Unaccounted NPS loads leading to unreliable load-reduction allocations or assimilation limits | |
Uncertainty about the future | Climate change | • Models accounting for climate change by incorporating results from climate predictions robust to such change • Medium- to long-term forecasting • Useful for critical condition modeling in an uncertain future | • Models not accounting for climate change not useful for medium- to long-term forecasting • Not useful for critical condition modeling in an uncertain future |
Sea-level rise | • Estuarine and coastal models driven by ocean models robust to sea-level rise • Models capable of evaluating BMP options useful for planning for the future | Steady-state models not useful to predict water levels, salinity, sediment transport, and aquatic vegetation in the future or evaluate BMP alternatives | |
LULC change | Models with dynamic LULC layers, or with inputs from LULC predictions incorporating changes in PS and NPS loads | Steady-state models based on existing GIS layers unable to account for changes in loading in the future due to LULC changes | |
Socioeconomic change | Models with flexible load limits and boundary conditions suitable for exploring changing loads and load-reduction allocations | Models with rigid load ceilings (e.g., legacy codes) infeasible when load sources and magnitudes change | |
Policy change | • Models with built-in error margin reliable even when policy changes • Increased stakeholder confidence • Model selection process with buffers for funding lapses and staff reductions, among other considerations | • Models specific to a particular policy not reliable when significant policy changes occur • Potential for loss of support for model development and maintenance |
Note: BMP = best management practice; GIS = geographic information system; LULC = land use/land cover; NPS = nonpoint source; and PS = point source.
Management Constraints
Principal constraints | Examples | Positive impacts | Negative impacts |
---|---|---|---|
Resource constraints | Funding | • Development of well-calibrated and confirmed models supported by data • Synoptic data collection and careful model development | Constrained choice of TMDL models to inexpensive ones that may not produce well-justified load-reduction allocations or assimilation limits |
Expertise | • Using data appropriately and developing strong synoptic data collection workflows • Building detailed high-quality process-based models | • Incorrect use of data resulting in representation errors and high uncertainty in model results • Inappropriate models resulting in large systematic errors and biases • Poorly motivated MOS and alternative scenario development | |
Labor | • Robust data collection • Rapid scheduling of tasks, prototyping and model development • Cross-functional teams | • Overexerted workforce prone to mistakes and poor task management • Time and budget overlays • Suboptimal data usage | |
Transparency | • Minimizing process loops and optimizing resource utilization • Robust knowledge-transfer • Culture of engagement, collaboration, mentorship, and skill-transfer | • Suboptimal process workflows with poorly informed stakeholder engagement • Miscommunications and information siloing • Working with outdated or erroneous data and methods and poorly maintained models • Stakeholder apprehensions about TMDL | |
Time | Use of robust decision process, calibration, and confirmation workflows and continuous stakeholder engagement resulting in high-quality TMDL models | Stringent time crunches due to litigation or other drivers result in inferior quality TMDL models | |
Stakeholder collaboration | Engagement | • Thoroughly vetted models and community buy-in with early and continuous stakeholder engagement • Robust developer and user groups for active exchange of ideas • Domain expertise and awareness of watershed characteristics and implementation constraints | • Apprehension and resentment toward TMDL when engagement does not occur frequently and in a conciliatory manner • Poorly reviewed models and lack of awareness about models |
Community participation | • Augmentation of data collection by citizen science and self-reporting by polluters • Synergistic cost-sharing by local PS dischargers and water consumers | Lack of community support leading to difficulty in gauging impact of TMDL | |
Scope of the TMDL model | Model use | • Scalable and user-friendly models allowing evaluation of changes in water quality criteria, beneficial uses, and exploration of alternative management scenarios • Planning for BMPs in stormwater and water quality management | • Unstable and resource-intensive models not suitable for decision support • Inability to evaluate role of BMPs, resulting in poorly designed load allocations and assimilation limits |
State-of-the-science | • Interim management actions by understanding current science • Time allocation for science to develop | • Inappropriate application of complex models • Loss of confidence in models | |
Regional issues | • Success of TMDL driven by consideration of local processes and drivers • Well-vetted and data-driven modeling in litigious environments | • Models developed for eastern and midwestern systems in the US inapplicable elsewhere due to more narrowly distributed flow ranges in the East and Midwest • Excessive litigation results in model development mired in controversy | |
Model complexity | Model parsimony for defense and clarity | Indefensible and poorly generalizing models due to unsupported model complexity |
Note: BMP = best management practice; MOS = margin of safety; and PS = point source.
Model Selection
Types of TMDL Models
Optimal Model Formulation
Choice of TMDL Model
Model Inputs and Outputs
Discussion
Conclusions
Data Availability Statement
Acknowledgments
References
Information & Authors
Information
Published In
Copyright
History
Authors
Metrics & Citations
Metrics
Citations
Download citation
Cited by
- Vamsi Krishna Sridharan, Saurav Kumar, Swetha Madhur Kumar, Can Remote Sensing Fill the United States’ Monitoring Gap for Watershed Management?, Water, 10.3390/w14131985, 14, 13, (1985), (2022).
- Charitha Gunawardana, Walter McDonald, A Socio-Economic Perspective on TMDL Development Progress, World Environmental and Water Resources Congress 2022, 10.1061/9780784484258.113, (1221-1231), (2022).
- Vamsi K. Sridharan, Steven C. McCutcheon, Nigel W. T. Quinn, Harry X. Zhang, Saurav Kumar, Ebrahim Ahmadisharaf, Xing Fang, Andrew Parker, Ebrahim Ahmadisharaf, Rene A. Camacho-Rincon, Xiaobo Chao, Xing Fang, William H. Frost, Mohamed M. Hantush, Sanaz Imen, Seshadri S. Iyer, Saurav Kumar, Rosanna J. La Plante, R. Craig Lott, Steven C. McCutcheon, Yusuf M. Mohamoud, Andrew Parker, Vamsi K. Sridharan, Zhonglong Zhang, Model Selection and Applications for Total Maximum Daily Load Development, Total Maximum Daily Load Development and Implementation, 10.1061/9780784415948.ch11, (319-356), (2022).
- Nigel W.T. Quinn, Steven C. McCutcheon, Vamsi K. Sridharan, Rosanna J. La Plante, Harry X. Zhang, Deva K. Borah, G. Padmanabhan, Ebrahim Ahmadisharaf, Rene A. Camacho-Rincon, Xiaobo Chao, Xing Fang, William H. Frost, Mohamed M. Hantush, Sanaz Imen, Seshadri S. Iyer, Saurav Kumar, Rosanna J. La Plante, R. Craig Lott, Steven C. McCutcheon, Yusuf M. Mohamoud, Andrew Parker, Vamsi K. Sridharan, Zhonglong Zhang, Introduction, Total Maximum Daily Load Development and Implementation, 10.1061/9780784415948.ch1, (1-30), (2022).