Grantee Research Project Results
Final Report: Development of a Surface Water Object-Oriented Modeling System (SWOOMS) for the Neuse River Estuary, North Carolina
EPA Grant Number: R827957Title: Development of a Surface Water Object-Oriented Modeling System (SWOOMS) for the Neuse River Estuary, North Carolina
Investigators: Luettich Jr., Richard A. , Bowen, J. , Alperin, Marc
Institution: University of North Carolina at Chapel Hill , University of North Carolina at Charlotte
EPA Project Officer: Packard, Benjamin H
Project Period: October 1, 1999 through September 30, 2002
Project Amount: $897,859
RFA: Computing Technology for Ecosystem Modeling (1999) RFA Text | Recipients Lists
Research Category: Environmental Statistics
Objective:
The Neuse River Estuary (NRE) shows many symptoms of an over-productive estuary: algal blooms, bottom water anoxia, and fish kills. Like most coastal ecosystems, primary production in the NRE is generally limited by nitrogen. Anthropogenic nitrogen loading to the Neuse River Basin has more than doubled since 1960, because of rapid growth in population and intensive agriculture in the watershed, increased atmospheric inputs, and groundwater inputs. Although the link between increased nitrogen loading and phytoplankton biomass within the estuary is difficult to establish, there is the widespread perception that impaired water quality within the NRE is related to cultural eutrophication. In the late 1990s, the North Carolina Department of Environment and Natural Resources was required under the Clean Water Act to develop a Total Maximum Daily Load (TMDL) for nitrogen in the Neuse River.
The overall objective of this research project was to improve modeling capabilities to allow better understanding of the ecological response of the NRE (and other estuarine systems) to varying nutrient loading and consequently to allow improved decisionmaking capabilities for determining such things as TMDLs. During the course of this project, we have worked to develop components that could fit into a multimedia modeling system that would ultimately allow researchers to track nutrients from their sources in the airshed and watershed to their arrival, storage, and impact in the estuary. Our specific efforts have been focused on developing the estuarine part of such a modeling system.
The specific objectives of this research project were to: (1) develop a finite-volume hydrodynamic model and transport model component; (2) develop a water column biogeochemical model component; (3) develop a sediment diagenesis model component; (4) evaluate the efficacy of implementing model components using an object-oriented framework; and (5) design and implement an object-oriented surface water model for seamless input/output with object-oriented models of the atmosphere, watershed, and groundwater media.
Summary/Accomplishments (Outputs/Outcomes):
Objective 1: Develop a Finite-Volume Hydrodynamic Model and Transport Model Component
Estuarine systems such as the NRE tend to have complicated geometries and bathymetries and are therefore ill suited for modeling with traditional techniques using regular, structured grids. The most common irregular, unstructured grid modeling technique is the Galerkin finite element method. Although this approach works well for purely hydrodynamic studies, it does not conserve mass on a grid cell by grid cell basis and is therefore poorly suited for solving highly nonlinear water quality transport equations. Consequently, in the present study, we proposed to develop hydrodynamic and transport model components using a finite volume methodology that readily allows for the use of irregular, unstructured grids and enforces strict mass conservation grid cell by grid cell. Although several low-order finite volume techniques have previously been applied to water quality problems, we chose to investigate the use of a new, hybrid finite volume/finite element approach, the Discontinuous Galerkin (DG) finite element method. The main advantages of this approach are its ability to handle advection-dominated flows, its strict cell by cell mass conservation properties, and the ease with which higher order interpolation and grid refinement can be handled. Recent applications of the DG method to transport problems have been very successful, although this method had not previously been applied to solve the shallow water hydrodynamic equations.
In the present study, we have developed and evaluated the DG methodology for the two-dimensional, vertically integrated shallow water hydrodynamic equations. Extension to the scalar transport equations is straightforward and presently in progress. Further expansion to three-dimensions also is underway. Our implementation makes use of approximating polynomials of arbitrary order over triangular elements that are linked via a numerical flux that is approximated using Roe’s method. Computational efficiency is maximized for high-order implementations by using an orthogonal and hierarchical triangular basis and efficient, symmetrical Gaussian quadrature rules that are near optimal for triangles. This results in a matrix-free algorithm with a minimum number of quadrature points.
Model performance is evaluated on several numerical test problems ranging from realistic tidally forced harbor problems to theoretical dam break problems. These test cases demonstrate that the model performs well for a range of hydrodynamic conditions, without the introduction of spurious oscillations or excessive damping of the solution. Additionally, systematic h (grid) and p (polynomial order) refinement establish the convergence of the method.
We believe that the DG method holds great promise for estuarine hydrodynamic/transport/water quality models because of its superior numerical solution properties and its strict cell-by-cell conservation. The primary disadvantage of the DG method at present appears to be a greater computational requirement for a given spatial discretization when compared to more traditional finite difference or finite element methods. We are continuing to evaluate potential computational improvements in the DG method as well as the trade off between the computational effort and the quality of the numerical solution.
Objective 2: Develop a Water Column Biogeochemical Model Component
As part of the development and testing of object-oriented water quality models, both a linked reservoir model and a one-dimensional river model were developed using procedural and object-oriented modeling paradigms. Simple biogeochemical models were incorporated into each of the object-oriented models. The river model has three water quality state variables (tracer, biochemical oxygen demand [BOD], and dissolved oxygen [DO]), whereas the reservoir model has five water quality state variables (tracer, dissolved solids, temperature, BOD, and DO). Compared to the river model, the object-oriented linked reservoir model, which was written in Java, uses a relatively large number of classes and a compartmentalized approach to define the physical system. This approach was taken to expedite later extension of the model to additional state variables and spatial dimensions. For instance, the linked reservoir spatial discretization scheme is functionally identical to the finite volume method and should therefore allow for a relatively straightforward extension to two or three dimensions. In its present form, however, the model is capable of modeling BOD, DO, and temperature dynamics in estuaries that can be schematized using the linked reservoir approach.
Two additional avenues of development were pursued using an existing water quality model (the Neuse Estuary Eutrophication Model [NEEM]). This model was an integral part of the NRE nitrogen TMDL analysis that was completed by one of the project principal investigators (J. Bowen) in collaboration with the North Carolina Department of Environment and Natural Resources in 2001. The water column biogeochemical model enhancements developed in the present project were used in this TMDL analysis.
One of these developments added the capability of running the NEEM in a coupled mode using a separate sediment diagenesis model, as might be done when a suite of environmental models is used for simulation. To run in coupled mode, the NEEM was modified to produce data files that could be used as input to the sediment diagenesis model. The sediment diagenesis model then calculated sediment/water column mass fluxes that were in turn read into the modified NEEM and used as an alternative to the calculations of its existing sediment diagenesis model. These mass fluxes were then used by the NEEM to calculate water column state variable concentrations. In the first use of the coupling, the models were iteratively run in “loose coupling” mode, in which the models were run consecutively, using the results from a previous run of the other model. Eventually, a coupling framework was created that allowed for a true collaborative and coupled use of the two models.
The other area of development in the NEEM was in the biogeochemical model itself. One potential challenge in coupling models of the airshed, the watershed, and the estuary is the staggering complexity of each of these models. As part of this project, we investigated the extent to which the predictive capability of the estuary model could be maintained while reducing the number of model parameters. In the existing model, three separate phytoplankton functional groups are used, and each of these groups requires 18 model parameters to define a phytoplankton group’s growth rate as a function of temperature, light, and nutrient concentration. Here, we attempted model simplification using sensitivity analysis and numerical optimization techniques. By eliminating two phytoplankton groups and utilizing a sensitivity analysis, we eliminated all but 7 of the original 54 model parameters. Not surprisingly, model error, as quantified by root mean square error between chlorophyll-a model predictions and observations, increased by approximately 75 percent over the original model. Next, an automated optimization technique was implemented to find the values of the remaining parameters that gave the lowest model error. The optimization procedure proved to be computationally burdensome, but produced promising results. Model error was reduced by approximately 60 percent, which brought the error of the simplified model close to that of the original. Because of the computational burden of the procedure, we were only able to run this test for 18 months of the full 42-month data set that was available, but it did provide some promise that substantial simplification in the estuary’s biogeochemical model could be achieved without sacrificing much in the way of prediction capability.
Objective 3: Develop a Sediment Diagenesis Model Component
Quantitative predictions regarding the impacts of mitigation on environmental systems are necessary, but elusive, owing to complex, nonlinear interactions between multiple reservoirs. This is especially true for shallow, poorly flushed estuaries such as the NRE. In this system, sediment-water column interactions provide an important control on cycling of limiting nutrients and oxygen demand. Therefore, we have developed, calibrated, and applied a sediment diagenesis model that simulates physical and biogeochemical interactions between the sediment and the water column. The sediment model is designed to be coupled to a water-column biogeochemical model, but also can function as a stand-alone modeling environment. Emphasis is placed on capturing the details of the sediment water interface—in particular, the diffusive boundary layer—to accurately simulate the impact of currents and concentrations in the bottom water on sediment processes and benthic fluxes. We report model results illustrating how sediment pools and processes respond to: (1) reduced organic matter deposition resulting from reduced nitrogen loading; (2) sudden changes in bottom water oxygen concentration; and (3) variable bottom currents.
The diagenetic model is based on a set of coupled, nonlinear differential equations that describe physical transport within the benthic boundary layer and surfacial sediments, and chemical and biological processes associated with sedimentary organic matter degradation. Species included in the model are: sedimentary organic matter (3-types), oxygen, nitrate, ammonium, sulfate, methane, total inorganic carbon, and two aggregated pools (dissolved and particulate) of reduced inorganic compounds produced directly or indirectly via anaerobic degradation. Transport processes include: molecular diffusion, eddy diffusion, bioturbation, and sedimentation. Model output includes sediment oxygen demand, benthic fluxes of nitrate and ammonium, concentration profiles of all chemical species, and depth distributions of denitrification, nitrification, sulfate reduction, methane production, and methane oxidation rates.
The sediment diagenesis model identifies a number of positive and negative feedbacks that modulate the estuarine response to changes in nitrogen loading. For example, the response of sediment oxygen demand to reduced organic matter deposition is tempered by slower respiration rates allowing deeper oxygen penetration and more efficient oxidation of ammonium and other reduced compounds. Hence, a 30 percent reduction in organic matter deposition translates to only a 20 percent reduction in benthic oxygen flux. On the other hand, the benthic ammonium flux is severely limited by even modest reductions in organic matter deposition. As a result, the net effect of reduced organic deposition is a positive feedback: a 30 percent reduction in organic matter deposition reduces the flux of recycling nitrogen from sediments by 40 percent.
In addition, the sediment model provides characteristic response times of the sediment compartment to changes in master variables such as organic matter deposition. For example, sediment oxygen demand responds very slowly (timescale: 10 years) to a reduction in organic matter deposition, owing to the large reservoir of reactive organic matter stored in sediments. In contrast, sediment processes respond to changes in bottom water oxygen and velocity on timescales of minutes.
Objective 4: Evaluate the Efficacy of Implementing Model Components Using an Object-Oriented Framework
One of the primary issues that motivated this study was the perception that existing estuarine water quality models have not been designed to be part of a larger modeling suite (e.g., including the airshed, watershed, and groundwater) and therefore may be poorly suited for linking with dynamic models of other environmental media. Water quality models have traditionally been written using procedurally based programming languages such as FORTRAN or C. In this type of programming, the model’s functionality is conceived as a collection of quantitative procedures that are each translated into computer code. Although this type of programming is conceptually simple, it can produce code that is difficult to share, extend, and reuse. For these and other reasons, a significant portion of commercial software is now written using a different programming paradigm. Object-oriented programs, written using languages such as Java, C++, or C# focus not on procedures, but on defining the attributes, methods, and relationships of distinct classes of objects. Collectively, these objects define the tasks to be completed or the physical system being modeled.
Within this project, we explored model design and implementation in the context of the unsteady, one-dimensional water flow and transport/water quality equations. Models were developed using coupled governing equations for mass and momentum (St. Venant equations) that were solved using an implicit time-stepping numerical scheme that allowed for flexibility in choosing time steps without concern for numerical instability. Replicate models were constructed using a procedurally based programming language (C) and two object-oriented programming languages (Java, C#). All three of these models were validated successfully by comparing their results to analytical results or to the results of test cases using previously validated models such as the U.S. Environmental Protection Agency-supported water quality model QUAL2E.
The development of models that shared governing equations and numerical formulations but differed in programming paradigm allowed for interesting comparisons between the procedural and object-oriented approaches. One such comparison concerned the reusability and ease of maintenance of the code. Although object-oriented codes are supposedly superior in these regards compared to procedural codes, our models written in C and Java differed little in reusability, extensibility, or ease of maintenance. From this work, it can be concluded that reusable and easily maintained codes can be developed using either approach if good programming practices are followed. Another expectation was that object-oriented programs are much slower to run as compared to procedurally based programs. Some significant speed differences were found between the two object-oriented programs, with the C# code being significantly faster than the Java code, but overall the object-oriented C# model was the fastest of the three. It must be pointed out, however, that the object-oriented C# code developed for this project had a very small number of classes, and therefore had a program structure that was not much different from a procedural code. A multidimensional model with a larger number of constituents would likely have many more classes and a more computationally expensive solution scheme. In this case, a procedural-based code may have a computational advantage. This proved to be the case for our linked reservoir model, as run times for the Fortran model were only 20 percent of the corresponding Java model.
Although the C# object-oriented code executed with surprising speed, where this programming approach really shined was in the easy integration of the model into the ArcGIS environment. ArcObjects programming was employed to incorporate the C# object-oriented model (QUALC#) into ArcGIS 8.2 using the Microsoft Visual Studio .NET framework. Inputs to the model such as boundary condition data, application specific physical data, and kinetic parameters and the output of the model results utilized a personal geodatabase that was designed specifically for the model. It was found that the object-oriented nature of the program made it relatively easy to incorporate into the ArcGIS environment. With this integration came a straightforward method for creating all the necessary input data, an easy means for running the program, and access to the vast array of post-processing capabilities of ArcGIS. This easy integration into a GIS framework and the capabilities the GIS environment offers may prove to be a decided advantage to surface water model developers that choose to use an object-oriented programming language for their models, and may drive the development of these sorts of models in the future.
Objective 5: Design and Implement an Object-Oriented Surface Water Model for Seamless Input/Output With Object-Oriented Models of the Atmosphere, Watershed, and Groundwater Media
This final objective represents a synthesis of all of the project components described above together with an integration of the estuarine modeling system with comparable modeling systems developed for other media. We did not achieve this final objective in the present project in part because of the time and effort required to complete the other objectives in this project, in part because we learned during the course of work that the object-oriented paradigm was not necessarily a modeling panacea, and in part because coordinated and holistic modeling efforts in the other media never materialized in such a way that facilitated development of a comprehensive multimedia modeling system. Consequently, this objective remains to be achieved, if it indeed remains as highly desirable as initially anticipated. Although linking models of different physical media would provide a valuable tool for tracking nutrients and managing water quality in estuarine and coastal systems, it is unclear whether a compelling reason exists to choose an object-oriented versus a procedural modeling structure for this effort.
Journal Articles on this Report : 6 Displayed | Download in RIS Format
Other project views: | All 44 publications | 8 publications in selected types | All 6 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Atkinson JH, Westerink JJ, Luettich Jr. RA. Two-dimensional dispersion analyses of finite element approximations to the shallow water equations. International Journal for Numerical Methods in Fluids 2004;45(7):715-749. |
R827957 (Final) |
Exit Exit |
|
Borsuk ME, Stow CA, Luettich Jr. RA, Paerl HW, Pinckney JL. Modelling oxygen dynamics in an intermittently stratified estuary: estimation of process rates using field data. Estuarine, Coastal and Shelf Science 2001;52(1):33-49. |
R827957 (Final) R825243 (1999) R825243 (Final) R826938 (2001) R826938 (Final) |
Exit Exit Exit |
|
Bowen JD, Hieronymus JW. A CE-QUAL-W2 model of Neuse Estuary for total maximum daily load development. Journal of Water Resources Planning and Management-ASCE 2003;129(4):283-294. |
R827957 (Final) |
Exit |
|
Buzzelli CP, Luettich Jr. RA, Powers SP, Peterson CH, McNinch JE, Pinckney JL, Paerl HW. Estimating the spatial extent of bottom-water hypoxia and habitat degradation in a shallow estuary. Marine Ecology Progress Series 2002;230:103-112. |
R827957 (Final) R826938 (2000) R826938 (Final) R828677 (2001) R828677C001 (Final) |
Exit Exit |
|
Reynolds-Fleming JV, Luettich Jr. RA. Wind-driven lateral variability in a partially mixed estuary. Estuarine, Coastal and Shelf Science 2004;60(3):395-407. |
R827957 (Final) R826938 (Final) R828677C001 (2003) |
Exit Exit Exit |
|
Stow CA, Roessler C, Borsuk ME, Bowen JD, Reckhow KH. Comparison of estuarine water quality models for total maximum daily load development in Neuse River Estuary. Journal of Water Resources Planning and Management-ASCE 2003;129(4):307-314. |
R827957 (Final) |
Exit Exit |
Supplemental Keywords:
marine, estuary, nitrogen oxides, modeling, water quality, nutrients, nitrogen, numerical methods, object-oriented, QUAL2E, multimedia, interdisciplinary, finite volume, discontinuous Galerkin finite elements, shallow water equation, transport, boundary layers, diffusion, sediment-water column exchange, hypoxia, anoxia, Neuse River Estuary, NRE, hydrodynamic model, transport model, North Carolina, NC,, RFA, Scientific Discipline, Water, Waste, Geographic Area, Ecosystem Protection/Environmental Exposure & Risk, Nutrients, Hydrology, Contaminated Sediments, Environmental Chemistry, Chemistry, State, Microbiology, computing technology, ecosystem modeling, fate and transport, aquatic ecosystem, environmental monitoring, nutrient supply, nutrient transport, aquatic modeling, watersheds, contaminated sediment, community decision making, environmental decision making, surface water object-oriented modeling system, surface water, wetland mitigation banking program, water quality, North Carolina (NC), information technology, biogeochemistry, cross-media environmental monitoring, groundwater, stream ecosystemRelevant Websites:
http://www.coe.uncc.edu/~jdbowen/swooms Exit
http://www.coe.uncc.edu/~jdbowen/neem Exit
http://citeseer.ist.psu.edu/herington03deco.html Exit
Progress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.