Journal of Water Resources Planning and Management

Successful engagement of stakeholders and scientists is essential when regulatory agencies want public input to environmental policy decisions that are technically complex. Standards of good practice for public involvement in regulatory decision making derived from the literature were used to evaluate stakeholder interactions with water quality models and modelers in the Neuse River total maximum daily load (TMDL) process. Some aspects of this interaction went well: stakeholders were involved early in the development of at least some of the models; presentations from modelers enhanced stakeholders’ scientific knowledge; and the regulatory agencies accepted stakeholder input as having a strong advisory role in their decisions. Some elements did not go so well: delays in delivery of modeling results prevented their full consideration by stakeholders; and incompatible timeframes for data collection, modeling, and regulatory decision making diminished confidence in the integrated modeling and monitoring framework that had been developed for the Neuse. Most importantly, the TMDL regulatory process was defined much too narrowly to encompass the stakeholders’ wide-ranging concerns for equity, cost effectiveness, and deliberate adjudication of the tradeoffs between costs and benefits of water quality regulation. Water quality monitoring and modeling emphasizing concentrations of chlorophyll did not begin to address the social, economic, and cultural concerns of the stakeholders. The most serious shortcomings of the Neuse TMDL process thus rest not with the scientists or the stakeholders, but with the too narrow structure of the regulatory process itself.

We develop a probability network model to characterize eutrophication in the Neuse River Estuary, North Carolina, and support the estimation of a total maximum daily load (TMDL) for nitrogen. Unlike conventional simulation models, probability networks describe probabilistic dependencies among system variables rather than substance mass balances. Full networks are decomposable into smaller submodels, with structure and quantification that reflect relevant theory, judgment, and/or observation. Model predictions are expressed probabilistically, which supports consideration of frequency-based water quality standards and explicit estimation of the TMDL margin of safety. For the Neuse Estuary TMDL application, the probability network can be used to predict compliance with the dissolved oxygen and chlorophyll a regulatory criteria as a function of riverine nitrogen load. In addition, the model includes ecological endpoints, such as fishkills and shellfish survival, that are typically more meaningful to stakeholders than conventional water quality characteristics. Incorporating these unregulated attributes into TMDL decisions will require explicit consideration of costs, benefits, and relative likelihoods of various possible outcomes under alternate loading scenarios.

In support of a total maximum daily load (TMDL) analysis, hydrodynamic and water quality conditions in the Neuse River Estuary were simulated for a 43 month period beginning in June 1996 using the two-dimensional, laterally averaged model CE-QUAL-W2. Several modifications to the model were made for this application including the further development of the existing sediment diagenesis submodel and an empirically derived light extinction prediction. The simulation period included two extreme nitrogen loading events and three summers with chlorophyll a concentrations that exceeded the water quality standard of 40 μg/L. A multiobjective, statistically based procedure was used to calibrate the model to observed chlorophyll a concentrations. The model was then used to predict changes in chlorophyll a concentrations and water quality standard exceedances in estuary regions for a variety of total nitrogen load reductions. The model predicted that a load reduction of approximately 5% was needed to lower the frequency of water quality standard violations below 10%. The process-based modeling approach was found to be a valuable tool for a TMDL analysis, with certain reservations related to its relatively high-data needs and its inability to fully address prediction uncertainty.

The Neuse River Estuary was included on the North Carolina Department of Water Quality 303(d) list for nutrients, and was scheduled for total maximum daily load (TMDL) development by spring of 2001. The water quality target of the TMDL was determined by the state to be the chlorophyll-a concentrations in the estuary. EPA Region 4 partnered with the State of North Carolina to provide technical assistance and guidance for nutrient TMDL development in the Neuse River Estuary. The goal was the development of hydrodynamic and water quality models that are sufficient to simulate the complex circulation and water quality kinetics within the system, including salinity and temperature stratification, wind driven seiching, dissolved oxygen stratification, and longitudinal and lateral variations in nutrient and chlorophyll-a concentrations. A three-dimensional, hydrodynamic, and water quality model was developed in the estuary from Maw Point at the Pamlico Sound boundary, to upstream at Streets Ferry Bridge above New Bern, North Carolina. The complex three-dimensional hydrodynamics of the Neuse estuary were modeled using the Environmental Fluid Dynamics Code (EFDC). EFDC was applied with water surface elevation forcing at the downstream boundary, freshwater inflows at the upstream boundaries, and wind over the water surface of the modeled domain. Water surface elevation, flows, currents, salinity, and temperature were simulated using EFDC. The U.S. EPA Water Quality Analysis Simulation Program (WASP6) was applied for the water quality portion of the model. The eutrophication model of WASP was used to simulate the complex nutrient transport and cycling in the estuary. The purpose of the water quality model is to predict a response in chlorophyll-a and dissolved oxygen concentrations as a function of nutrient loadings and transport throughout the Neuse River Estuary. The model was used to evaluate various loading scenarios and the impact on water quality within the “use support” areas within the 303(d) listed segments. The hydrodynamic and water quality models were calibrated for 1998 and confirmed for 1999 and 2000. A comparison of the model simulations with the extensive dataset shows that the models are accurately simulating the longitudinal/seasonal distribution of the hydrodynamics, mass transport, and water quality. The water quality model was used to evaluate TMDL scenarios.

The North Carolina Division of Water Quality developed a total maximum daily load (TMDL) to reduce nitrogen inputs into the Neuse River Estuary to address the problem of repeated violations of the ambient chlorophyll a criterion. Three distinct water quality models were applied to support the TMDL: a two-dimensional laterally averaged model, a three-dimensional model, and a probability (Bayesian network) model. In this paper, we compare the salient features of all three models and present the results of a verification exercise in which each calibrated model was used to predict estuarine chlorophyll a concentrations for the year 2000. We present six summary statistics to relate the model predictions to the observed chlorophyll values: (1) the correlation coefficient; (2) the average error; (3) the average absolute error; (4) the root mean squared error; (5) the reliability index; and (6) the modeling efficiency. Additionally, we examined each model’s ability to predict how frequently the 40 μg/L chlorophyll a criterion was exceeded. The results indicate that none of the models predicted chlorophyll concentrations particularly well. Predictive accuracy was no better in the more process-oriented, spatially detailed models than in the aggregate probabilistic model. Our relative inability to predict accurately, even in well-studied, data-rich systems underscores the need for adaptive management, in which management actions are recognized as whole-ecosystem experiments providing additional data and information to better understand and predict system behavior.

The spatially referenced regressions on watershed attributes modeling approach, as applied to predictions of total nitrogen flux in three North Carolina river basins, addresses several information needs identified by a National Research Council evaluation of the total maximum daily load program. The model provides reach-level predictions of the probability of exceeding water-quality criteria, and estimates of total nitrogen budgets. Model estimates of point- and diffuse-source contributions and nitrogen loss rates in streams and reservoirs compared moderately well with literature estimates. Maps of reach-level predictions of nutrient inputs and delivery provide an intuitive and spatially detailed summary of the origins and fate of nutrients within a basin.

Assessment of water quality conditions as required by Section 303d of the Clean Water Act must rely on limited monitoring data. Because data are limited there will always be the potential for error when deciding if water quality standards are being met. However, the informed use of statistical procedures makes it possible to describe and manage these errors. We make the case for using such procedures in water quality assessment and draw the implications for the monitoring data for water quality standard setting, arguing that standard setting must accommodate the limits of the monitoring data and the statistical procedures that will be used for water quality assessment.

A simplified framework for considering variability and uncertainty in developing lake phosphorus total maximum daily loads (TMDLs) is demonstrated. Explicit consideration of these factors can increase the probability that TMDL implementation will meet a defined water quality goal at an acceptable frequency. Although a lake goal is typically expressed as a seasonal or yearly average phosphorus concentration, effects of temporal variations can be captured by correlating average phosphorus concentrations with the frequency of algal blooms (defined by extreme values of chlorophyll a) or the frequency of exceeding numeric water quality standards that are directly linked to algal blooms, such as hydrogen ion concentration or transparency. The margin of safety (MOS) required to achieve the lake goal at a defined frequency and with a defined confidence level can be estimated by including stochastic terms in the phosphorus balance equation to reflect variability and uncertainty. Given limitations in the data and models typically used in developing TMDLs, the cost of the MOS, expressed in terms of percent of the total allocated load or safety factors in the design of control measures, can be large. The MOS would be expected to increase with the percent load reduction required under the TMDL, as the forecast loads become increasingly dependent on assumptions regarding the performance of best management practices or other measures for reducing loads. The magnitude and cost of the MOS can be reduced by implementing TMDLs in an iterative fashion with ongoing data collection, and model refinement to reduce uncertainties associated with forecasting the performance of phosphorus load controls and lake responses.

We present a computational approach for identifying the significance of uncertainty in assessing the consequences of sediment and nutrient transport in a section of the Chattahoochee River south of Lake Lanier, as it passes through Atlanta, Georgia. Specifically, our analysis aims at identifying the key control and management actions, and the key scientific uncertainties about the fluvial system, that govern the attainment of a set of water quality objectives for the downstream boundary of the study area. To this end, we present a computational framework that integrates a recently developed sediment-nutrient dynamics model with a Monte Carlo-based methodology for model uncertainty evaluation. Our results suggest that, in general, reliable execution of controls and management actions is more crucial to meeting the target values for flow, sediment, and phosphorus concentration, than the scientific uncertainties associated with fluvial processes within the river channel. We also discuss the potential utility of our framework for accommodating the various science- and policy-derived uncertainties in the total maximum daily load process.