Ecohydrology Component of Louisiana\'s 2012 Coastal Master Plan: Mass-Balance Compartment Model

Journal of Coastal Research

SI

67

16–28

Coconut Creek, Florida

Summer 2013

Ecohydrology Component of Louisiana’s 2012 Coastal Master Plan: Mass-Balance Compartment Model Ehab Meselhe†, John A. McCorquodale‡, Jeff Shelden§, Mark Dortch§, T. Stokka Brown§, Peter Elkan§, Mallory D. Rodrigue††, Jennifer K. Schindler††, and Zhanxian Wang§ †

The Water Institute of the Gulf 301 N. Main Street, Suite 2000 Baton Rouge, LA 70825, U.S.A. [email protected]

‡

Pontchartrain Institute for Environmental Sciences FMI Center for Environmental Modeling University of New Orleans New Orleans, LA 70148, U.S.A.

§

Moffatt & Nichol 1616 East Millbrook Road, Suite 160 Raleigh, NC 27609, U.S.A.

††

Fenstermaker 1100 Poydras Street, Suite 1550 New Orleans, LA 70163, U.S.A.

ABSTRACT Meselhe, E.; McCorquodale, J.A.; Shelden, J.; Dortch, M.; Brown, T.S.; Elkan, P.; Rodrigue, M.D.; Schindler, J.K., and Wang, Z., 2013. Ecohydrology component of Louisiana’s 2012 Coastal Master Plan: mass-balance compartment model. In: Peyronnin, N. and Reed, D. (eds.), Louisiana’s 2012 Coastal Master Plan Technical Analysis, Journal of Coastal Research, Special Issue No. 67, 16–28. Coconut Creek (Florida), ISSN 0749-0208. Coastal Louisiana is a complex system that encompasses large expanses of wetlands interspersed with shallow bays and estuaries of varying sizes and degrees of connectivity to the Gulf of Mexico, numerous water control structures, large riverine systems, and an intricate system of natural and manmade channels. This complex system is experiencing devastating rates of land loss that have been exacerbated by subsidence and sea level rise. As part of Louisiana’s 2012 Coastal Master Plan, this modeling effort utilizes an efficient mass-balance approach to provide coastwide (~100,000 km2), long-term (~50 y) performance projections for proposed restoration and protection measures. The model presented here provided detailed information about the spatial and temporal variability of water depth, salinity, accretion rates, deposition, and other water quality parameters across the Louisiana coastal zone. Furthermore, the model provided this information to subsequent modules in the master plan suite of models, namely, wetland morphology, vegetation, ecosystem services, and barrier shoreline morphology. Collectively, this suite of models served as an effective approach to provide valuable comparative assessments for the various proposed restoration and protection scenarios and alternatives.

ADDITIONAL INDEX WORDS: Modeling, coastal, hydrology, water quality, estuaries, accretion, erosion, sediment.

INTRODUCTION Louisiana’s Comprehensive Master Plan for a Sustainable Coast (2012) was initiated by the Coastal Protection and Restoration Authority (CPRA) of Louisiana to evaluate the performance of potential protection and restoration projects on the Louisiana Coast for the next 50 years (CPRA, 2012). The numerical modeling effort presented in this paper is one component of an integrated set of models that provides a projected ‘‘outlook’’ of the coast under various future scenarios and project actions (Peyronnin et al., 2013). This projected outlook with any of the actions can then be compared to a similar outlook without the action, referred to herein as future without action (FWOA). The scenarios modeled included moderate and less optimistic natural conditions reflected through variability in rainfall, evapotranspiration (ET), sea level rise, subsidence, and DOI: 10.2112/SI_67_2.1 received 26 October 2012; accepted in revision 15 February 2013; corrected proofs received 7 May 2013. Ó Coastal Education & Research Foundation 2013

upstream freshwater runoff (for specifics of the scenarios, see Peyronnin et al., 2013). The output of the ecohydrology model is utilized as input in the models for wetland morphology (Couvillion et al., 2013), vegetation (Visser et al., 2013), ecosystem services (Nyman et al., 2013), and barrier islands and shoreline morphology (Hughes et al., 2012). The ecohydrology modeling effort relied primarily on the mass-balance approach. Previously developed models were used as the starting point, such as the Chenier Plain hydrological model (Meselhe, McCorquodale, and Georgiou, 2008) and the water quality model for the Pontchartrain Estuary (McCorquodale et al., 2008, 2009; Roblin, 2008). Massbalance models have been used in other coastal wetland applications (Habib et al., 2007, 2008; Meselhe, Waldon, and Arceneaux, 2010) and have proved to be effective and computationally efficient tools. Although such models do not account for the full dynamic of coastal processes, they can capture the dominant long-term and large-scale spatial trends necessary for planning purposes. The following section provides a brief discussion of the model selection process. The

Mass-Balance Compartment Model

17

Figure 1. The inset shows a map of the states surrounding the northern Gulf of Mexico, while the box outlines the study area shown in the large image. The large image is a map of the study region, with each modeling region outlined and labeled.

model setup, calibration, and validation, as well as the modeling scenarios, are all discussed in subsequent sections.

MODEL SELECTION The selection of a suitable modeling approach is governed by several factors: the objectives of the study, the temporal and spatial resolution of the required output, the spatial extent of the study area, the temporal duration of the simulations, the number of scenarios and alternatives, and the available time and budget to conduct the study. For the modeling effort presented herein, the spatial extent of the Louisiana coastal area covers an area of nearly 100,000 km2. One of the objectives of the study is to investigate long-term morphological changes to the coastal area. As such, 50-year simulations were required and nearly 400 protection and restoration measures were evaluated and assessed. In addition, the models presented in this paper provided information to other modeling components (wetland morphology, vegetation, and ecosystem service models) as part of the comprehensive analysis to support coastal master plan development. Given these constraints, it was critical to select a modeling approach with a reasonable run time. As such, it was determined that a mass balance for

constituents, a balance of gravity and bottom friction for flow, or a simplified conservation of momentum (the diffusion wave) is a reasonable and computationally efficient approach. It also provided a suitable level of resolution in outputs for this planning-level analysis.

MODEL DESIGN Compartment Structure Modeling efforts were divided among three regions of the Louisiana coast: Lake Pontchartrain–Barataria Basin (PB), Atchafalaya Basin (AA), and Chenier Plain (CP; Figure 1). A mass-balance, link-node, compartment model was developed in the Formula Translating (Fortran) system programming language by the University of New Orleans (Meselhe, McCorquodale, and Georgiou, 2008; Meselhe and Miller, 2008). The first link-node model was developed in the 1960s by Water Resource Engineers Inc. and was published as the Dynamic Estuary Model by the Federal Water Quality Administration (Feigner and Harris, 1970). Since then, the concept has been used in the U.S. Environmental Protection Agency (EPA) Storm Water Management Model (EPA, 2006) and Water

Journal of Coastal Research, Special Issue No. 67, 2013

18

Meselhe et al.

Quality Analysis Simulation Program (EPA, 2010) models. The model has been used in multiple projects since then, including the Coastal Louisiana Ecosystem Assessment and Restoration (CLEAR) Water Quality Modeling of the Pontchartrain Estuary (McCorquodale et al., 2008; Roblin, 2008), the National Oceanic and Atmospheric Administration (NOAA) Mississippi River Gulf Outlet Violet Optimization Study (Georgiou et al., 2009), and a joint project for the EPA, NOAA, CLEAR, and Lake Pontchartrain Basin Foundation to study Pontchartrain Estuary salinity, nutrients, and sediment dynamics (McCorquodale et al., 2009). Because the 2012 Coastal Master Plan modeling effort called for coastwide hydrodynamic modeling, the PB Fortran model was used as the basis for setting up the Berkeley Madonna (Macey and Oster, 2001) equation solver for the AA and CP regions (herein referred to as the PB, AA, and CP models, respectively). The mass-balance approach was used to analyze state variables, and the governing equations and assumptions were similar for all three regions. Parameters were computed using hybrid upwinding (Ferziger and Peric, 1996), along with the central difference numerical integration scheme in the PB model and the upwind scheme in the AA and CP models (Versteeg and Malalasekera, 2007). In the PB model, the area inside of each compartment was separated into upland, marsh, and open water subcompartments with links providing exchange between subcompartments (Figures 2a and b). Open water subcompartments represented bays, lakes, and other water bodies that are hydraulically connected within the domain, while links represented channels, passes, and waterways that connect these open water bodies. Local riverine systems, including the Mississippi system, supplied freshwater to the open water subcompartments. Upland subcompartments exchanged water, sediment, and nutrients with the marsh subcompartments via runoff. The offshore boundary was composed of individual nodes with links to the nearest model compartments. In the AA and CP models, compartments were separated into distinct types (channel, open water, and marsh) with links providing exchange between compartments (Figures 2c and d). Local rivers provided freshwater to the compartments. The AA offshore boundary was composed of compartments, whereas the CP offshore boundary was composed of individual nodes. Because governing equations were not applied to the boundary, the use of either nodes or compartments at the offshore boundary was acceptable and had no effect on the outputs of the internal compartments. To account for the water inside a marsh compartment, a percentage of land was applied to each compartment, where 100% indicated the compartment was entirely land or marsh and 0% indicated the compartment was entirely water. Table 1 shows the number of compartments and surface types for each model region. In the PB model, runoff from upland and marsh areas was conveyed to open water subcompartments. Direct river inflow and flows and loads from diversion projects were treated as inputs to the open water compartments. In addition, wastewater nutrient loads and discharge were modeled in a similar approach as any riverine inflow. Runoff from upland areas in the AA model and local runoff in the CP model were conveyed to

the marsh and open water compartments. Discharges and nutrient loads from local wastewater treatment plants were modeled as local riverine inflow in the AA and CP models.

Governing Equations Equations used in the models were based on theoretical concepts found in the literature on hydrodynamic processes (e.g. discharge; ET; precipitation, or P; volume; stage; depth; and structure schedules) and water quality processes (e.g. nutrient cycles and nitrogen removal, salinity, total suspended solids and accretion, chlorophyll-a as algae, and water age) found in nature. These equations were computed sequentially within each compartment for all state variables at a predetermined time step.

Hydrodynamic Processes In the hydrodynamic modeling, it was assumed that the flows were governed by gravity and friction force. Discharges modeled included exchange flows, riverine flows, distributary or diversion flows, atmospheric contributions, and runoff contributions. Wastewater inflows were solely used to improve the calibration of the water quality models; therefore, wastewater discharges were excluded from the hydrodynamic modeling. Riverine flows, distributary or diversion flows, atmospheric inputs, and runoff discharges were defined as boundary conditions. Values of ET were assumed to be the same in every model simulation year. At every time step, the exchange flows of the links Qi between compartments were calculated using Equation (1). Equation (1) is derived from a balance of the difference in energy between the compartments and the loss of energy due to friction and eddy losses in the interconnecting links shown as yellow arrows on Figure 2 (Chow, 1959): 8  1   9 2 > > d SAL i SALj > X  ð3600  24Þ Qi ¼ Ai > > > > i : ; kim þ 2gn2i L4=3 Ri

ð1Þ where i is the link identifier, j is the upstream compartment identifier, j þ 1 is the downstream compartment identifier, and Qi is the water flow rate in link i (in cubic meters per day). In addition, Ai is the cross-sectional water flow area for link i (in square meters), g is the gravitational constant (9.81 m/s2), Ej is the water surface elevation of compartment j (in meters), di is the centroidal water depth for link i (in meters), SALj is the concentration of salinity in compartment j (in parts per thousand or kilograms per cubic meter), kim is the minor loss coefficient for link, ni is the Manning’s roughness coefficient, Li is the length for link i (in meters), and Ri is the hydraulic radius for link i (in meters). The cross-sectional water flow area for link i (Ai) was the product of the flow width for link i (wi) and the flow depth of link i (Hi). The flow width was obtained by approximating the width in which water can pass between two compartments. In the CP and AA models, the flow depth was defined as the minimum depth of the two connecting compartments. The PB model used the average depth, rather than the minimum depth, of flow in each link at every time step. For links

Journal of Coastal Research, Special Issue No. 67, 2013

Mass-Balance Compartment Model

19

Figure 2. Model dynamics for the PB, AA, and CP models.

connected to an offshore compartment (AA) or node (PB and CP), the depth was defined as the depth of the upstream or nearshore compartment (assuming that the Gulf of Mexico boundary was always downstream). The centroidal depth was the depth at the centroid, or geometric center, of the water area (wet area in a compartment). The minor loss coefficient was a calibration parameter to account for link geometry that cannot be captured using the standard attributes. The Manning’s roughness coefficient for a link (n) was the average of the roughness values of the

Table 1. Number of compartments per surface type in each ecohydrology region. Region

Channel

Open Water

Marsh

Upland

Offshore

AA CP PB

73 28 183 links

21 19 —

67 102 89

— — —

4 6 nodes 7 nodes

* PB compartments contain subcompartments for marsh, open water, and upland. AA and CP models designate a compartment for each surface type.

two connected compartments. The roughness of the offshore links in the CP and PB models was designated as the roughness of the upstream or nearshore compartment. The roughness of the offshore links in the AA model was the average roughness of the two connected compartments. The length of a link was defined as the approximate travel distance between the centroids of the two connected compartments. The hydraulic radius was defined as the volume divided by the wetted area of the link. The direction of flow was based on the defined upstream and downstream compartments and the resulting sign of the value between the absolute symbols as computed with Equation (1). The new stage (water surface elevation) in compartment j (Ej) in Equation (1) is obtained by adding the change in depth from Equation (2) to the previous stage. A minimum water depth was defined to eliminate undefined answers in the exchange flow equation. The PB and AA models required a minimum depth of 0.01 m. The CP model required minimum depths for wetting and drying. The minimum depth for the drying, or emptying, of a compartment was called the

Journal of Coastal Research, Special Issue No. 67, 2013

20

Meselhe et al.

‘‘dry’’ depth (Hdry) and equaled 0.01 m. A compartment was considered dry (i.e. containing 0% water), when it reached this depth. This was simulated in the model by restricting the outflows in the compartments when depth equaled Hdry. To avoid instability, a ‘‘wet’’ depth (Hwet) of 0.02 m was applied during the wetting of the compartments, meaning that a compartment was restricted from releasing water until reaching Hwet and above. The PB and AA models did not use a wet or dry depth, because stability was maintained during the wetting and drying processes. The change in volume of the water within each compartment (Vj, in cubic meters) depends on the change in the stage or depth. The volume was estimated from the discharges mentioned previously, as well as the atmospheric exchange of P and ET. Equation (2) is the continuity relationship and defines the change in depth as a function of exchange flows and atmospheric exchanges: X X X X ! Qj;trib þ Qj;div þ Qj;run Qj;i þ dHj ¼ dt Asj þ ðPj  ETj Þ

ð2Þ

where dHj/dt is the rate of change in water depth for compartment j (in meters per day), Qj is the flow to compartment j from all links i (in cubic meters per day), Qj,trib is the flow to compartment j from all tributaries (in cubic meters per day), Qj,div is the flow to compartment j from all diversions and distributaries (in cubic meters per day), Qj,run is the flow to compartment j from all runoff contributions (in cubic meters per day), Pj is the P on compartment j (in meters per day), ETj is the ET from compartment j (in meters per day), and Asj is the water surface area of compartment j (in square meters). The initial depth was estimated from the compartment bed elevation and water stage from the initial conditions. For the CP channel compartments, the depth was defined for each compartment using a depth–volume relationship curve and the calculated volume.

Water Quality Processes The water quality component of the models included processes that govern the transport and reactions of conventional water quality variables that either are dissolved or are in particulate form in the water column. The models also included processes that deposit or transfer material onto the sediment bed, but they did not predict the fate of those constituents once deposited. The water column was assumed to be fully mixed and aerobic at all locations and times. Thus, there was no transfer of nutrients from the bed to the water column under anoxic, or oxygen-deprived, conditions. The mass-balance– based, reactive transport equation for each water quality constituent can be stated as follows: X 0 Ck; j;s Qj;s kdis X Ai Ck; j Vj dCk; j ¼ þ  ðCk; j  Ck; jn Þ dt Vj Vj Vj i L i X Ssk; j; l 1000Lk Asj þ ð3Þ þ Vj Vj where dCk,j/dt is the rate of change of concentration of constituent k in compartment j (in grams per cubic meter or milligrams per liter); s is the water source via a tributary, a

Table 2. Water quality constituents or state variables simulated in the models. Parameter (unit)

Abbreviation

Frequency

Stage (m, NAVD88) Total suspended solids (mg/L) Accretion (from inorganic TSS, g/m2/y) Salinity (ppt) Tidal range (m, NAVD88) Total Kjeldahl nitrogen (mg/L) Water temperature (8C) Nitrate þ nitrite nitrogen (mg/L) Ammonium nitrogen (mg/L) Dissolved organic nitrogen (mg/L) Total phosphorus (mg/L) Soluble inorganic phosphorus (mg/L) Phytoplankton as chlorophyll-a (lg/L) Detritus (mg/L) Residence times (water age, days) Nitrogen removal (denitrification, g/m2/y)

STG TSS ACC SAL TRG TKN TMP NO3 NH4 DON TPH SPH ALG DET AGE NRM

Daily Monthly Yearly Monthly Monthly Monthly Monthly Monthly Monthly Monthly Monthly Monthly Monthly Monthly Monthly Yearly

NAVD88 ¼ North American Vertical Datum of 1988.

diversion or distributary, or runoff; Ck,j,i,s is the concentration of constituent k flowing into or out of compartment j via link i or a source s (in grams per cubic meter or milligrams per liter); Qi,s is the water flow rate entering or exiting compartment j via link i or a source s (in cubic meters per day); Ai is the cross-sectional water flow area for link i (in square meters); Vj is the water volume of compartment j (in cubic meters); Vj 0 is the change in water volume with respect to time for compartment j (cubic meters per day); t is the time (in days); kdis is the dispersion coefficient (in square meters per day); Li is the length of link i (in meters); Ck,jn is the concentration of constituent k in compartment jn adjacent to compartment j (in grams per cubic meter or milligrams per liter); Ssk,j,l is the rate of change of mass of constituent k in compartment j due to source/sink l associated with kinetic processes including transformations or reactions and settling (in grams per day ); Asj is the water surface area for compartment j (in square meters); and Lk is the regional atmospheric deposition rate (for water compartments) or local loading flux (e.g. marsh or wetland delivery) for constituent k (in kilograms per square meter per day). The sources or sinks and reaction terms for each of the state variables were based primarily on the formulations used in the QUAL2K water quality model (Chapra and Pelletier, 2003). Equation (3) was solved within each computational compartment, regardless of compartment type. Table 2 describes the various water quality constituents or state variables simulated in the models. Dissolved oxygen (DO), biochemical oxygen demand, organic carbon, and related DO source or sink processes were not included, because these models assume that the water column was always aerobic. The aerobic assumption was used to avoid a two-layer model; this implies that there are no hypoxic or anoxic zones that would preclude denitrification. To overcome this deficit in the model, a calibrated denitrification rate was introduced to represent the anaerobic layer at the bed.

Model Assumptions Several assumptions were made to successfully model the hydrodynamics and water quality of the coastal domains:

Journal of Coastal Research, Special Issue No. 67, 2013

Mass-Balance Compartment Model

(1) All variables changed with respect to time, but they were all uniform in space within a compartment. (2) The cross-sections of the links were represented by an equivalent rectangular shape. (3) The calibrated roughness coefficients and the diffusivities were constants for each link. (4) The wind shear on the links was assumed to be negligible. (5) The sediment resuspension was based on the Krone (1962) approach: The critical shear was related to the critical wind speed squared, and similarly the bed shear was assumed to be proportional to the mean daily wind speed squared. (6) A reference settling velocity of 8 m/d was adjusted for salinity and water temperature. (7) The focus of the model simulations was on long-term trends. As such, short-term (e.g. hours) hydrological events were assumed to be negligible. (8) The results were integrated spatially over large compartments and then temporally averaged over a day. Under these circumstances, the momentum exchange (temporal and spatial acceleration) did not play a primary role. Even if it had been taken into account, its impact would be lost in the temporal and spatial averages. Therefore, the momentum exchange between channels and open water, lakes, or marshes was assumed negligible. (9) In the PB model, sediment accretion was assumed to occur by separate processes in the marsh and the open water. The depths in the links were adjusted based on the bed elevations in the subtending open water cells. (10) Sediment accretion in the channel compartments was assumed to be zero because it was difficult to represent the narrow canals in the 30 3 30 m digital elevation model. This assumption did not affect the capacity of the channels and their ability to convey flow. (11) To ensure continued ability to convey flow, a deposition reduction factor was applied in the AA model to limit deposition and prevent compartments from completely filling with sediment. In the PB model, the link crosssectional areas were adjusted to maintain a peak channel forming velocity (~1.2 m/s). If accretion resulted in excess velocities in a link, it was assumed that the delta formation would propagate and downstream channels would receive more sediment. (12) The benthic layer was not explicitly modeled. It was assumed that these processes could be represented by temperature-dependent rate constants. For example, denitrification was represented by a temperature-dependent denitrification rate coefficient. (13) It was assumed that the open water nutrient processes were more important than benthic layer processes.

Boundary Conditions Boundary conditions (composed of atmospheric, riverine, distributary or diversion, offshore, wastewater, and runoff data) were collected from several observation stations throughout the model domains. Atmospheric boundary conditions were generated using observed P, ET or evaporation, and wind data. Riverine and distributary or diversion conditions were ob-

21

tained from observed discharge, water quality, temperature and salinity data, or a combination of these, while wastewater and runoff boundary conditions were obtained from observed discharge and water quality data. Offshore boundary conditions were obtained from tidal stage, salinity, water quality, and temperature data. In the PB model, a temperature boundary was applied to the offshore boundary nodes, the Mississippi River, and its distributaries and diversions. A temperature boundary was defined for each tributary in the AA and CP regions. See the ecohydrology appendix to the 2012 Coastal Master Plan (Meselhe et al., 2012) for agency names and dates used for each parameter. All available data in the period 1990–2009 were compiled across the three ecohydrology regions and was used to create a 25-year dataset for the model boundaries. If data records were incomplete or unavailable, the modeling teams used statistical filling techniques for each boundary condition type. The three teams worked independently and followed similar procedures when possible, but data collection was determined by each team based on variations in data availability and the schedule mandated by the 2012 Coastal Master Plan.

Initial Conditions The initial conditions of the model were determined based on the best temporal and spatial data available for 1990 (which was assumed to be year 0 in the simulation). For the PB model, a ramp-up of 1 day was used in the simulations to assimilate initial conditions to acceptable values. For the AA model, the initial conditions for all parameters were set to zero; thus, the first month of the simulation acted as a spin-up period. For the CP region, a model spin-up was used for determining initial conditions, where parameters were set to zero and run for a month. Then the resulting month was averaged (except for the first day) to acquire a new initial value. The required initial conditions for each compartment included the values for total suspended solids, nitrate þ nitrite nitrogen, ammonium nitrogen, dissolved organic nitrogen, dissolved organic phosphorus, soluble phosphorus, detritus, phytoplankton as chlorophyll-a, salinity, stage, and residence time or water age. See the ecohydrology model appendix to the 2012 Coastal Master Plan (Meselhe et al., 2012) for agency names and dates used for each parameter.

MODEL TESTING, CALIBRATION, AND VALIDATION Evaluation Tests While it is impossible to prove that there were no coding or logical errors in the models, a series of evaluation tests were performed during the model development. These tests were intended to verify that (1) the models responded properly to the boundary conditions, (2) the models were mass conserving, and (3) the models reproduced spatial and temporal trends observed in the field measurements. The tests required manipulation of the river inflows and atmospheric boundary conditions, as well as the application of a simplistic hydrodynamic boundary such as an exaggerated tidal wave. A series of six tests were performed on the models to ensure that the model responds to a simplified input and reproduces a logical output,

Journal of Coastal Research, Special Issue No. 67, 2013

22

Meselhe et al.

Table 3. Stage and parameter values for the ecohydrology model hydrodynamic tests. Stage (m) NAVD88 Test No. 1 2 3 4 5 6

(step) (square (square (square (square (square

wave) wave) wave) wave) wave)

Day 0–2

Day 2–4

Day 4–6

Day 6–8

Day 8–10

Day . 8

Inflows (m3/s)

P (mm/d)

ET (mm/d)

5 5 0 5 5 0

5 6 1 5 6 1

5 5 0 5 5 0

5 4 1 5 4 1

5 5 0 5 5 0

5 5 0 5 5 0

0 0 0 0 0 0

0 0 0 ET ET ET

0 0 0 ET ET ET

NAVD88 ¼ North American Vertical Datum of 1988.

such as the general increase of all stages in response to an idealized step change in the stage at the open boundary. It is a validity check on the hydraulic equations in the code. Table 3 shows the step values in the open boundary stage for six intervals of 2 days. The models passed the six tests described in Table 3. Compartment stages increased and decreased according to the imposed boundary signal and then returned to the constant boundary stage. Figure 3 graphically illustrates one such test (test 5). In this test, all compartments in the model experienced an initial stage of 5.0 m and were subjected to a tidal signal that increased from 5.0 to 6.0 m, returned to 5.0 m, decreased to 4.0 m, and returned to 5.0 m, where it remained constant until

equilibrium was achieved. Input of ET and P were equal, and there were no tributary inflows. A lag in reaching the maximum stage was visible at interior compartments, because it took time for the wave to propagate inland.

Calibration and Validation Each model was calibrated through the adjustment of flow parameters, reaction parameters, and boundary conditions. The models were subsequently validated using another period from the available records. The PB model was calibrated using the period of 2008 and later years and validated using measured data using the period of 2007 and earlier years. The AA model was calibrated using observed field values for the 2007 calendar year; it was then validated with observed

Figure 3. Evaluation test 5 simulating a simple boundary wave with P equivalent to ET. The black line represents the base-case conditions, while grayscale lines represent the stage response of randomly selected compartments.

Journal of Coastal Research, Special Issue No. 67, 2013

Mass-Balance Compartment Model

23

(1) The models were designed to calculate mean representative values of variables for large compartments and were compared to measured gauge data values from specific locations. (2) The model calibration was based on the simulated daily average, which reduced effects due to tidal variations when compared to an instantaneous measurement. (3) A large uncertainty in bathymetric data and measured stage data existed. (4) There was minimal boundary condition data forcing in the models (i.e. offshore salinity and water quality). (5) In a large-scale link-node model such as this, the hydrological connections were not fully represented in detail. Instead, a general approximation of the connectivity was utilized. (6) There was a lack of wind forcing on large interior open water compartments. (7) The models were unable to reproduce stratification, because cells were modeled as a fully mixed column. For comprehensive model results and further discussion on the assessment of the model performance, see the ecohydrology technical appendix (Meselhe et al., 2012).

RESTORATION PROJECT SCENARIOS Contribution to Planning Effort Figure 4. Stage (A) calibration and (B) validation. The solid line represents model results; the dotted line represents observed measurements.

field values for the calendar years 2008 and 2009. The CP model was calibrated using observed field values for the 2007 calendar year; it was then validated with observed field values for the 2010 calendar year. Calibration and validation years were chosen based on data availability (daily continuous data being preferred). Several observation stations across the Louisiana coastal area provide intermittent discrete measurements. As a result, statistical techniques were used to fill in missing records when necessary. See the ecohydrology appendix to the 2012 Coastal Master Plan (Meselhe et al., 2012) for agency names and dates used for each parameter. Overall, the model calibration and validation results agreed fairly well with the measurements. Because this effort is part of a planning-level analysis (Peyronnin et al., 2013), graphical inspection of the model results was sufficient in assessing the models’ agreement with observed measurements. Examples of calibration and validation results for stage and salinity are illustrated in Figures 4 and 5, respectively. More plots of calibration and validation results can be found in the ecohydrology appendix to the 2012 Coastal Master Plan (Meselhe et al., 2012).

Assessment of Model Performance Models were assessed based on their ability to replicate observed measurements. Discrepancies between simulated and observed values resulted from the following factors:

As a contribution to the 2012 Coastal Master Plan modeling effort, the ecohydrology models provided essential information on the hydrodynamics and quality of water in the coastal regions of Louisiana. The models were run for two sets of 25 years; a transference of output occurred after each 25-year period. For example, after the first 25 simulation years, the ecohydrology stage, salinity, and sediment output were used as input into the wetland morphology model. Subsequently, the wetland morphology land use and elevation output were used as input to the ecohydrology models. At the end of simulation (year 50), an outlook was produced that allowed adequate planning-level screening of proposed projects. In addition, multiple modeling scenarios were simulated to estimate the range of model sensitivity to external parameters. More details on the transference process and the sensitivity scenarios can be found in the master plan overview article in this special issue (Peyronnin et al., 2013).

Project Descriptions and Implementation The master plan modeling was divided into two phases: the project-level phase and the group-level phase. In the projectlevel phase, individual projects were simulated to estimate their total impact on all water quality parameters in the local area. In the group-level phase, individual projects were grouped together to estimate their combined impact on the entire coast. Projects in the project-level phase were selected with the protection and restoration of the Louisiana coast as the primary criteria. Hurricane protection projects were included in the group-level phase. The following paragraphs describe how these projects were incorporated into the ecohydrology models.

Journal of Coastal Research, Special Issue No. 67, 2013

24

Meselhe et al.

Figure 5. Salinity calibration and validation. The black line represents model results; the gray and white markers represent observed measurements.

Marsh Creation Projects Marsh creation projects create new wetlands in open water areas by mechanically placing dredged material. Land elevation and percentage of land for each compartment were updated based on outputs from the wetland morphology model.

Hydrological Restoration Projects Hydrological restoration project features are designed to restore natural hydrological patterns either by conveying freshwater to areas that have been cut off by manmade features or by preventing the intrusion of saltwater into fresh areas through manmade channels and eroded wetlands. To model these projects, links and channels were created as needed. These additional links were dimensioned based on proposed channel depth and width. In addition, and as needed, existing link depth and width were modified, flow was adjusted based on control structure operation, and land areas were increased.

Ridge Restoration Projects Ridge restoration focuses on the reestablishment of historic ridges in basins. Projects often affected flow links between compartments. When the ridge was to be constructed inside a compartment boundary and the ridge height was less than the channel depth, the Manning’s n for the link was increased.

Finally, compartment elevations and percentage of land were based on outputs from the wetland morphology model.

Barrier Island Restoration Projects Barrier island restoration projects create and restore dune, beach, and back-barrier marsh in an effort to restore or augment Louisiana’s offshore barrier islands and headlands. Flow links in the vicinity of barrier islands were adjusted based on proposed barrier island configurations and the crosssectional area of tidal inlets provided by the barrier morphology model.

Sediment Diversion Projects Sediment diversion projects use channels, structures, or both to divert sediment and freshwater from the Mississippi and Atchafalaya Rivers into adjacent basins. To model diversion projects, the model team added supplemental inflows at boundaries and links between diversion intakes and release compartments or redistributed flows among existing links. Link capacities affected by diversions were increased to tolerate larger flow conditions. In the PB model, the percentages of flow to the existing passes and proposed diversions were based on Hydrologic Engineering Centers River Analysis System (HEC-RAS) modeling of the Mississippi River (Davis, 2010), and the percentages of sand in the suspended sediment were prorated to the river flow squared. For large diversions, it

Journal of Coastal Research, Special Issue No. 67, 2013

Mass-Balance Compartment Model

25

Figure 6. Simulations of mean annual accretion rates for two Lower Mississippi River sediment diversion options for the area in the red box at the top of the map. (A) Rates for a large sediment diversion in upper Breton Sound. (B) Rates for a large sediment diversion in lower Breton Sound. Diversion locations are represented by yellow stars.

was assumed that the water flows in a fanlike or deltaic pattern toward the Gulf of Mexico. In the AA model, the diverted flow was based on the flow condition in the water source location, and the diverted sediment concentration was assumed to be the same as the concentration of the source.

Oyster Barrier Reef Projects Bioengineered oyster reefs improve oyster propagation and serve as breakwaters to attenuate wave energies. To model these projects, the bed elevation of affected compartment links was adjusted to the height of the reef.

Hurricane Protection Projects

Channel Realignment Projects In channel realignment projects, channels and/or structures divert sediment and freshwater from the Mississippi and Atchafalaya Rivers into adjacent basins. For the PB model, Mississippi River flow and sediment load was distributed to nearby compartments in a manner similar to diversion projects and in a deltaic pattern. The capacity of links connecting nearby, downstream, or both compartments was increased to permit additional flow and encourage flow propagation toward the Gulf of Mexico. Downstream of the realignment, the flow distribution in the river was set to zero to simulate closure of the river. For the CP model, compartments and links were adjusted to replicate the new channel alignment.

Levees are one way to help reduce flood risk to communities. When proposed levees bisected compartments or links, the links were either cut off or reduced. Links were reduced based on the number of gates and their dimensions.

DISCUSSION OF RESULTS The ecohydrology models were efficient at simulating the impacts of protection and restoration alternatives over a 50year period, with the spatial scale extending more than 100,000 km2. The temporal changes could be resolved daily, weekly, or monthly, while the spatial resolution ranged from less than 1 to nearly 6000 km2 per compartment. The following summary describes the impacts of the protection and restoration projects

Journal of Coastal Research, Special Issue No. 67, 2013

26

Meselhe et al.

Figure 7. Monthly averaged salinity in the CP region of Louisiana for January 2051 for FWOA (A) and master plan conditions during drought conditions (less optimistic scenario; B).

proposed through the 2012 Coastal Master Plan on the various regions of the Louisiana coast. Land building is a primary objective in the master plan. The numerical simulations tested the ability to build land by diverting water and sediment from neighboring rivers. The numerical simulations focused on continuous operations reflecting a proportion of river flow with a minimum amount of flow retained in the river. At low river flows, diversions did not operate. The resources of these rivers (freshwater, sediment, energy, and nutrients) are limited and must be considered as a whole, i.e. as a system. When reallocation of these resources is planned, the models are valuable tools in simulating and evaluating the complex interactions among these resources.

Figure 6 demonstrates the impact that diversion size can have on accretion. It shows the effects of a single large upstream Mississippi River diversion (in upper Breton Sound; Figure 6A) and a single large downstream Mississippi River diversion (in lower Breton Sound; Figure 6B). A comparison of Figures 6A and B illustrates that when large quantities of flow (and sediment) are extracted upriver, less sediment is available for distribution downriver, e.g. at the mouth of the Mississippi River. This was evident in the Atchafalaya region as well. The analysis showed that diversions are an effective strategy of delivering sediment to a region. However, the conveyance of sediment accompanied by freshwater diversions results in the delivery of low-salinity water to the receiving water region. This may have other ecosystem effects. Ecosystem service

Journal of Coastal Research, Special Issue No. 67, 2013

Mass-Balance Compartment Model

models, fed by information from the ecohydrology model, assist in the evaluation of effects on the ecology of these receiving basins. A dependent relationship exists between accretion and salinity when large diversions are utilized. The upper and lower Breton Sound diversion projects illustrate this fact. With both large diversions, the increase in the accretion rate corresponds to a decrease in salinity in these areas. Marsh creation and restoration projects that reduce the influx of saltwater can reduce salinity as well. Changes in the land– water configuration within an estuary can promote higher accretion rates. In many areas of the Louisiana coast, the natural hydrology has been disrupted over the years, allowing saline waters from the Gulf of Mexico to propagate inland via channels and canals. This open access causes riverine freshwater to flow into the Gulf, thereby preventing distribution into the nearby marshes and lakes. By reducing the connectivity of inland areas to these tidal waters of offshore regions, the salinity would subsequently decrease. Control structures reduce the capacity of channels, thereby inhibiting tidal water intrusion and allowing upstream riverine freshwater and local storm runoff to be distributed through marsh areas. In areas connected to rivers, the compartments become less saline as the river water is allowed to permeate farther south. Marsh creation and ridge restoration projects can reduce the penetration of saline Gulf waters into the estuary. Finally, numerical simulations illustrated the strong potential impact of climate change (represented in hypothetical changes to the P and ET). Marsh areas that are primarily driven by the balance between P and ET showed a strong response (changes in water quality, especially salinity) to drought conditions (Figure 7). Such a response emphasizes the critical need for implementing a water management plan to conserve freshwater resources.

CONCLUSIONS The ecohydrology model was developed for the entire Louisiana coast and was calibrated and validated for the following state variables: stage, discharge, salinity, suspended solids, temperature, nutrients, and algae as chlorophyll-a. For modeling purposes, the coast was treated in three regions: PB, AA, and CP. The models for each region were based on interconnected, fully mixed compartments. The models computed the state variables at time steps on the order of minutes and then averaged these values at daily, weekly, monthly, and yearly scales. The flow distributions from the Mississippi River to the compartments in the PB model were determined using a calibrated HEC-RAS model (Davis, 2010). These validated models were run for FWOA and for master plan conditions. Projects that resulted in the greatest change in water quality conditions included sediment diversions, ridge and hydrological restoration, and marsh creation projects. The models provided state variable output (salinity and accretion being of primary importance) for each compartment for the FWOA and master plan conditions over a total simulation period of 50 years. Output from the ecohydrology model was provided to other modeling teams for use as input to subsequent models.

27

One of the important outcomes of this holistic modeling approach was the quantification of the relative responses of the system to project size and location. In the case of sediment diversions, the study showed that these projects can build land; however, the approach indicated that the Mississippi and Atchafalaya Rivers are finite resources in terms of energy, water, and sediment. Subsequently, there are tradeoffs when land–building diversions are introduced, e.g. an upriver diversion of energy, water, and sediment reduces the energy, water, and sediment at downriver distributaries and diversions. Similarly, ridge restoration and marsh creation projects can aid in sediment retention but may modify channel and tidal inlet area dimensions, thereby reducing the transport of water quality constituents.

ACKNOWLEDGMENTS The modeling effort presented in this manuscript was funded by the CPRA.

LITERATURE CITED Chapra, S.C. and Pelletier, G.J., 2003. QUAL2K: A Modeling Framework for Simulating River and Stream Water Quality: Documentation and Users Manual. Medford, Massachusetts: Civil and Environmental Engineering Department, Tufts University, 121p. Chow, V., 1959. Open-Channel Hydraulics. New York: McGraw-Hill, 680p. Couvillion, B.R.; Steyer, G.D.; Wang, H.; Beck, H.J., and Rybczyk, J.M., 2013. Forecasting the effects of coastal protection and restoration projects on wetland morphology in coastal Louisiana under multiple environmental uncertainty scenarios. In: Peyronnin, N. and Reed D.J. (eds.), Louisiana’s 2012 Coastal Master Plan Technical Analysis. Journal of Coastal Research, Special Issue No. 67, pp. 29–50. CPRA (Coastal Protection and Restoration Authority), 2012. Louisiana’s Comprehensive Master Plan for a Sustainable Coast. Baton Rouge, Louisiana: CPRA, 190p. http://www.coastalmasterplan.la. gov. Davis, M.D., 2010. Numerical Simulation of Unsteady Hydrodynamics in the Lower Mississippi River. New Orleans, Louisiana: University of New Orleans, Master’s thesis, 300p. EPA (U.S. Environmental Protection Agency), 2006. Storm Water Management Model (SWMM). http://www.epa.gov/nrmrl/wswrd/ wq/models/swmm. EPA (U.S. Environmental Protection Agency), 2010. Water Quality Analysis Simulation Program (WASP). http://www.epa.gov/athens/ wwqtsc/html/wasp.html. Feigner, K.D. and Harris, H.S., 1970. Documentation Report: FWQA Dynamic Estuary Model. Washington, DC: Florida Water Quality Association, U.S. Interior Department, 248p. Ferziger, J. H. and Peric, M., 1996. Computational methods for fluid dynamics. Berlin: Springer-Verlag, 364p. Georgiou, I.Y.; McCorquodale, J.A.; Schindler, J.; Retana, A.G.; FitzGerald, D.M.; Hughes, Z., and Howes, N., 2009. Impact of multiple freshwater diversions on the salinity distribution in the Pontchartrain Estuary under tidal forcing. In: FitzGerald, D.M. and Reed D.J. (eds.), Geologic and Environmental Dynamics of the Pontchartrain Basin. Journal of Coastal Research, Special Issue No. 54, pp. 59–70. Habib, E.; Larson, B.; Nuttle, W.K.; Nelson, B.; Rivera-Monroy, V.H.; Meselhe, E.A., and Twilley, R.R., 2008. Effect of rainfall spatial variability and sampling on salinity prediction in an estuarine system. Journal of Hydrology, 350(1–2), 56–67. Habib, E.; Nuttle, W.K.; Rivera-Monroy, V.H.; Gautum, S.; Meselhe, E.A., and Twilley, R.R., 2007. Assessing effects of data limitations on salinity forecasting in Barataria Basin, Louisiana, using a Bayesian analysis. Journal of Coastal Research, 23(3), 749–763.

Journal of Coastal Research, Special Issue No. 67, 2013

28

Meselhe et al.

Hughes, Z.; Weathers, D.; Georgiou, I.; FitzGerald, F., and Kulp, M., 2012. Appendix D-3: barrier shoreline morphology model technical report. In: Louisiana’s Comprehensive Master Plan for a Sustainable Coast. Baton Rouge, Louisiana: Coastal Protection and Restoration Authority, pp. 1–43. Krone, R.B., 1962. Flume Studies of the Transport of Sediment in Estuarial Processes. Final Report. Berkeley, California: Hydraulic Engineering Laboratory and Sanitary Engineering Research Laboratory, University of California, 110p. Macey, R.I. and Oster, G.O., 2001. Berkeley Madonna: Modeling and Analysis of Dynamic Systems. http://www.berkeleymadonna.com (accessed November 3, 2011). McCorquodale, J.A.; Georgiou, I.Y.; Retana, A.G., and Roblin, R.J., 2008. Assessment of Integrated Hydrodynamic and Transport for Long-Term Predictions, Volumes 1 and 2. Final Report. Baton Rouge, Louisiana: Coastal Louisiana Ecosystem Assessment and Restoration, 384p. McCorquodale, J.A.; Roblin, R.J.; Georgiou, I.Y., and Haralampides, K.A., 2009. Salinity, nutrient, and sediment dynamics in the Pontchartrain Estuary. In: FitzGerald, D.M. and Reed D.J. (eds.), Geologic and Environmental Dynamics of the Pontchartrain Basin. Journal of Coastal Research, Special Issue No. 54, pp. 71–87. Meselhe, E.A.; McCorquodale, J.A., and Georgiou, I.Y., 2008. Hydrologic Modeling and Budget Analysis of the Southwestern Louisiana Chenier Plain—Part III: Sediment Budget for the Chenier Plain, Louisiana. State of Louisiana Interagency Agreement 2503-06-16. Vicksburg, Mississippi: Louisiana Coastal Area (LCA) Science and Technology Program Office, 68p. Meselhe, E.A.; McCorquodale, J.A.; Shelden, J.; Dortch, M.; Brown, S.; Rodrigue M.; Wang, Z.; Elkan, P., and Schindler, J., 2012. Appendix D-1: eco-hydrology model technical report. In: Louisiana’s Comprehensive Master Plan for a Sustainable Coast. Baton Rouge, Louisiana: Coastal Protection and Restoration Authority, pp. 1–493.

Meselhe, E.A. and Miller, R.L., 2008. Hydrologic Modeling and Budget Analysis of the Southwestern Louisiana Chenier Plain— Part I: Regional Model. State of Louisiana Interagency Agreement 2503-06-16. Vicksburg, Mississippi: Louisiana Coastal Area (LCA) Science and Technology Program Office, 219p. Meselhe, E.A.; Waldon, M.G., and Arceneaux, J.C., 2010. Water budget model for a remnant northern Everglades wetland. Journal of Hydraulic Research, 48(1), 100–105. Nyman, J.A.; Baltz, D.M.; Green, M.; Kaller, M.D.; Leberg, P.L.; Richards, C.P.; Romaire, R.P., and Soniat, T.M., 2013. Likely changes in habitat quality for fish and wildlife in coastal Louisiana during the next 50 years. In: Peyronnin, N. and Reed, D.J. (eds.), Louisiana’s 2012 Coastal Master Plan Technical Analysis. Journal of Coastal Research, Special Issue No. 67, pp. 60–74. Peyronnin, N.; Green, M.; Parsons Richards, C.; Owens, A.; Reed, D.; Chamberlain, J.; Groves, D.G.; Rhinehart, K., and Belhadjali, K., 2013. Louisiana’s 2012 coastal master plan: overview of a sciencebased and publicly informed decision-making process. In: Peyronnin, N. and Reed, D.J. (eds.), Louisiana’s 2012 Coastal Master Plan Technical Analysis. Journal of Coastal Research, Special Issue No. 67, pp. 1–15. Roblin, R.J., 2008. Water Quality Modeling of Freshwater Diversions in the Pontchartrain Estuary. New Orleans, Louisiana: University of New Orleans, Master’s thesis, 220p. Versteeg, H.K. and Malalasekera, W., 2007. An Introduction to Computational Fluid Dynamics: The Finite Volume Method, 2nd edition. Harlow, England: Pearson Prentice Hall, 517p. Visser, J.M.; Duke-Sylvester, S.M.; Carter, J., and Broussard, W.P., III, 2013. A computer model to forecast wetland vegetation changes resulting from restoration and protection in coastal Louisiana. In: Peyronnin, N. and Reed, D.J. (eds.), Louisiana’s 2012 Coastal Master Plan Technical Analysis. Journal of Coastal Research, Special Issue No. 67, pp. 51–59.

Journal of Coastal Research, Special Issue No. 67, 2013