A wetland hydrology and water quality model ... - Wiley Online Library

WATER RESOURCES RESEARCH, VOL. 43, W04434, doi:10.1029/2006WR005003, 2007

A wetland hydrology and water quality model incorporating surface water/groundwater interactions Cevza Melek Kazezyılmaz-Alhan,1 Miguel A. Medina Jr.,2 and Curtis J. Richardson3 Received 28 February 2006; revised 7 September 2006; accepted 16 October 2006; published 28 April 2007.

[1] In the last two decades the beneficial aspects of constructed treatment wetlands have

been studied extensively. However, the importance of restored wetlands as a best management practice to improve the water quality of storm water runoff has only recently been appreciated. Furthermore, investigating surface water/groundwater interactions within wetlands is now acknowledged to be essential in order to better understand the effect of wetland hydrology on water quality. In this study, the development of a general comprehensive wetland model Wetland Solute Transport Dynamics (WETSAND) that has both surface flow and solute transport components is presented. The model incorporates surface water/groundwater interactions and accounts for upstream contributions from urbanized areas. The effect of restored wetlands on storm water runoff is also investigated by routing the overland flow through the wetland area, collecting the runoff within the stream, and transporting it to the receiving water using diffusion wave routing techniques. The computed velocity profiles are subsequently used to obtain water quality concentration distributions in wetland areas. The water quality component solves the advection-dispersion equation for several nitrogen and phosphorus constituents, and it also incorporates the surface water/groundwater interactions by including the incoming/outgoing mass due to the groundwater recharge/discharge. In addition, output from the Storm Water Management Model (SWMM5) is incorporated into this conceptual wetland model to simulate the runoff quantity and quality flowing into a wetland area from upstream urban sources. Additionally, the model can simulate a water control structure using storage routing principles and known stage-discharge spillway relationships. Citation: Kazezyılmaz-Alhan, C. M., M. A. Medina Jr., and C. J. Richardson (2007), A wetland hydrology and water quality model incorporating surface water/groundwater interactions, Water Resour. Res., 43, W04434, doi:10.1029/2006WR005003.

1. Introduction [2] Wetlands are important since they stabilize water supplies and clean polluted water as downstream receiving water bodies [Richardson, 1995; Mitsch and Gosselink, 2000]. Surface water/groundwater interactions occur in wetlands and pollutants in either surface or groundwater are mixed and the quality of both sources is affected by each other. Therefore, in wetland modeling, it is important to account for the effect of surface water/groundwater interactions. The role of wetlands as one of the best management practices to decrease the surface runoff peaks and to improve surface runoff water quality has recently been demonstrated [Richardson et al., 1997; Borin et al., 2001; Moore et al., 2002; Mitchell et al., 2002]. Consequently, there have been more research efforts aimed at understanding wetland hydrology and water quality. The 1 Department of Civil Engineering, Auburn University, Auburn, Alabama, USA. 2 Department of Civil and Environmental Engineering, Duke University, Durham, North Carolina, USA. 3 Duke University Wetland Center, Nicholas School of the Environment and Earth Sciences, Duke University, Durham, North Carolina, USA.

Copyright 2007 by the American Geophysical Union. 0043-1397/07/2006WR005003

current research efforts are on modeling wetland hydrology, modeling wetland water quality, and modeling surface water/groundwater interactions in wetlands. [3] The USGS modular three-dimensional finite difference groundwater flow model (MODFLOW) has been applied to different wetland sites by several investigators to simulate wetland hydrology [Bradley, 1996, 2002; Restrepo et al., 1998]. MODFLOW can simulate the spatial and temporal variations of the groundwater table and therefore is suitable mostly for wetland types where the water table is below the surface throughout the year and the water flows into the subsurface. For the wetland sites where MODFLOW is insufficient, extra computer packages for MODFLOW have been developed. For example, the South Florida Water Management District (SFWMD) has developed several MODFLOW packages in order to take into account South Florida hydrologic conditions and complexities [Restrepo et al., 1998]. For surface flow dominated wetlands, the models simulate variations in surface water depths [Hammer and Kadlec, 1986; McKillop et al., 1999; Lee et al., 2002; W. Walker and R. Kadlec, Dynamic Model for Stormwater Treatment Areas (DMSTA), http:// www.walker.net/dmsta/index2.htm]. These models predict the movement of surface waters; however, they do not incorporate a surface water/groundwater interaction compo-

W04434

1 of 16

W04434

KAZEZYıLMAZ-ALHAN ET AL.: WETLAND HYDROLOGY-WATER QUALITY MODEL

W04434

Figure 1. Hydrological components of the WETSAND model. nent (which might have a significant effect on wetland hydrology, especially because of the strategic location of wetlands where the groundwater is shallow). [4] The WETLAND model developed by Lee et al. [2002] and the dynamic model for storm water treatment areas (DMSTA) model developed by W. Walker and R. Kadlec (Dynamic Model for Stormwater Treatment Areas (DMSTA), http://www.walker.net/dmsta/index2.htm) for Everglades Water Conservation Areas also simulate water quality in wetlands in addition to wetland hydrology. Wetland water quality models have been developed in order to estimate nutrient retention and removal [Dorge, 1994; Martin and Reddy, 1997; Kadlec and Hammer, 1988]. Although these models simulate the reaction sequence of nitrogen dynamics and other nutrients, they do not account for the effect of surface water/groundwater interactions on nutrient transport. [5] The importance of modeling interactions between groundwater and wetlands and their effect on wetland functions are discussed in detail by Winter et al. [1998], Winter [1999], and Price and Wadington [2000]. Many investigators observed the effect of surface water/groundwater interactions on wetland hydrology and contaminant transport through experiments at different wetland sites [Winter and Rosenberry, 1995; Devito and Hill, 1997; Choi and Harvey, 2000; Warren et al., 2001]. Recently, Krasnostein and Oldham [2004] developed a conceptual model to describe the interactions between a wetland and local groundwater, and Keefe et al. [2004] used a solute transport model with transient storage to evaluate contaminant transport in wetlands.

[6] Although these studies are helpful in understanding the role of surface water/groundwater interactions in wetlands, there is a need for more dynamic and sophisticated models that account explicitly for the effect of surface water/ groundwater interactions. In many of these models, lumped parameter rather than distributed system components are used in calculating the water and chemical budgets. Whereas many site studies acknowledge the existence of surface water/groundwater interactions in wetlands, there are very few mathematical models which represent these interactions adequately. This study contributes to these research efforts by developing a general comprehensive dynamic wetland model (WETSAND), which has both water quantity and water quality components, and incorporates the effects of surface water/groundwater interaction. This model considers a wetland site as a distributed system, and therefore predicts the water levels and pollutant distributions not only as a function of time but also as a function of space throughout the system. The water quality component simulates the fate and transport of Phosphorus and Nitrogen compounds by incorporating the effect of mass exchange between the surface and subsurface water phases. WETSAND also takes into account the effect of flow generated from upstream areas. The restored wetland system at Duke University is selected as the application site to investigate the capabilities of this conceptual wetland model.

2. Model Development [7] The graphical representation of the conceptual wetland hydrological model is shown in Figure 1. During a

2 of 16

W04434


W04434

Figure 2. Major components of the WETSAND model. rainfall event, overland flow occurs over both urbanized watershed areas upstream and wetland areas downstream. The surface runoff and associated water quality constituents, which are generated over the urban areas, are connected to the wetland areas through the stream tributaries. Wetlands are located in the transitional zones between uplands and streams. They have a very mild slope (< % 0.1). In a storm event, the overland flow that occurs on upland sites is collected in the wetland areas and continues to flow into the streams. Water from both upstream and wetland sites generally flows downstream into receiving waters (e.g., lakes or estuaries). Surface water on a wetland is lost by vegetation transpiration, surface evaporation or by infiltration. The surface water/groundwater interaction depends on the groundwater table elevation. If the groundwater table is above the surface water level in a wetland then groundwater discharges; otherwise, groundwater recharges. [8] The wetland model (WETSAND) has two main components: (1) a wetland water quantity model and (2) a wetland water quality model. The structure of this general wetland model is shown in Figure 2. 2.1. Wetland Water Quantity [9] The diffusion wave equation is used to solve for flow over a wetland area because diffusion wave theory applies to the milder slopes (0.1% to 0.01%), for which kinematic wave theory is insufficient. This is important since wetlands usually are located on terrains with very mild slopes. The one-dimensional diffusion wave equation is written as follows: @y @y @2y þ c ¼ K1 2 þ q @t @x @x

diffusivity (L2/T) and q is the water source/sink term (L/T). Wave celerity c and hydraulic diffusivity K1 are defined as c ¼ mV

Vy 2S0

ð2Þ

where V is the water velocity (L/T), S0 is the bottom slope (L/ L) and m is the exponent of water depth that appears in the relationship between flow rate and water depth. The water quantity component of WETSAND computes the water depth and therefore the wetland hydraulics only in the longitudinal direction of the wetland site. Although inflow and outflow at many wetland sites have been observed in more than one direction, the main flow is in the direction of the decreasing slope of the terrain which allows us to use a one-dimensional formulation. [10] In order to calculate wetland overland flow, water depth (as given by equation (1)) needs to be related to flow rate. Kadlec et al. [1981] investigated the hydrology of overland flow in wetlands. They found that results obtained by using the Manning equation did not match measured data collected from a wetland site. The explanation lies in the fact that most flows in wetlands are usually transitional, somewhere between laminar and turbulent. Kadlec [1990] explained that calculation of overland flow in wetlands could be accomplished with an appropriate friction rule that is a power law for velocity in terms of depth and the friction slope. This rule reflects both the effect of a vertical vegetation stem density gradient and a bottom elevation distribution. Consequently, the flow rate on a wetland site is given by [Kadlec and Knight, 1996]:

ð1Þ

where y is the surface water depth (L), t is time (T), x is the distance (L), c is the wave celerity (L/T), K1 is the hydraulic

K1 ¼

Q¼

ð1 107 Þ Wy3 S0 ð5 107 Þ Wy3 S0

K¼

1 107 5 107

ð3Þ

for dense and sparse vegetation, respectively, where Q is the flow rate in (m3 d1), W is the wetland width (L), K is the

3 of 16


W04434

coefficient which reflects the vegetation density (m1 d1), and the exponent of y is m = 3. [11] The flow rate can also be written as Q ¼ VA ¼ VWy

)V ¼

ð1 107 Þ y2 ðS0 Þ ð5 107 Þ y2 ðS0 Þ

ð4Þ

[12] In diffusion wave theory, the term S0 is replaced by S0 @y/@x. Therefore the surface water velocity on a wetland area is calculated as follows: )V ¼

ð1 107 Þ y2 ðS0 @y=@xÞ ð5 107 Þ y2 ðS0 @y=@xÞ

ð5Þ

[13] The term q in equation (1) contains rainfall, groundwater discharge, and lateral inflow as water source terms and infiltration, evapotranspiration, and groundwater recharge as water sink terms: q ¼ qr qinf qet þ qdrch þ ql

ð6Þ

where qr is rainfall intensity (L/T), qinf is the infiltration rate (L/T), qet is evapotranspiration rate (L/T), qdrch is the rate of groundwater recharge/discharge (L/T) and ql is the lateral inflow term. [14] In order to calculate infiltration rate, a modified version of the Green-Ampt method during unsteady rainfall is used [Chu, 1978]. The original Green-Ampt equation was formulated to describe the infiltration process under a ponded surface [Green and Ampt, 1911] and subsequently improved by Mein and Larson [1973]. If there is no ponding, then all the rainfall infiltrates into the soil. For steady rain, infiltration starts with no ponding. Ponding begins later when the rainfall rate exceeds the infiltration rate during the rainfall event, which lasts until the rain stops. In an unsteady rain, there may be several periods when the rainfall intensity exceeds the infiltration rate. Therefore it is important to determine the time that ponding begins, which is provided by the modified Green-Ampt method. The formulation for the modified Green-Ampt method is as follows. [15] In case 1, no ponding occurs (ij < fj+1 < fj): ij fj

Fjþ1 ¼ Fj þ ij Dt

MS fjþ1 ¼ Kz 1 þ Fjþ1

ð7Þ

M ¼ qs qi

ð8Þ

if fj+1 ij ) fj = ij otherwise, case 2 applies. [16] In case 2, ponding begins to occur ( fj+1 < ij < fj): Kz MS Fp ¼ ij K

dt ¼

Fjþ1 ¼ Fp þ Kz ðDt dt Þ þ MS ln ) fj ¼

Fjþ1 Fj Dt

Fp Fj i

Fjþ1 þ MS Fp þ MS

ð9Þ

[17] In case 3, ponding occurs ( fj+1 < fj < ij): ð11Þ

ij > fj

Fjþ1 ¼ Fj þ Kz Dt þ MS ln ) fj ¼

Fjþ1 þ MS Fj þ MS

Fjþ1 Fj Dt

ð12Þ

where ij is the rainfall intensity at the jth time step (L/T), fj is the infiltration rate at the jth time step (L/T), Fj is the total infiltration at the jth time step (L), S is the suction head at wetting front (L), Kz is the vertical hydraulic conductivity (L/T), M is the moisture of water deficit, qi is the initial moisture content, qs is the saturated moisture content, Fp is the total infiltration when ponding begins (L) and dt is the time to the start of ponding (T). [18] The total amount of evapotranspiration per month is calculated using the Thornthwaite [1948] method. The term qet is calculated by dividing the total amount of evapotranspiration per month by the total number of time steps in a month. [19] The surface water/groundwater interaction is taken into account by groundwater recharge/discharge terms. The groundwater recharge or discharge, respectively, is calculated following Darcy’s law: qdrch

@H ¼ Kx @x

0

ð13Þ

where H is total head (L) and Kx is the horizontal hydraulic conductivity (L/T). The exchange between surface and groundwater is calculated in the lateral direction at the banks of the wetland. [20] Each term involved in q is variable in time. The change of rainfall intensity through time depends on the frequency of the recording time of the measured data at gauging stations. As infiltration is calculated by using the rainfall data, the time interval for infiltration rate will be the same as that for rainfall intensity. Evapotranspiration requires average monthly temperature and therefore the value of evapotranspiration rate changes each month. The groundwater recharge/discharge component requires groundwater level data and therefore the time interval for the rate of groundwater recharge/discharge depends upon the frequency of the recorded groundwater level data. Since each of these components may have a different timescale, the values are interpolated linearly based on the smallest time interval available among the terms. [21] The overland flow generated from upland areas is also calculated using diffusion wave theory. The difference between the overland flow occurring on wetlands and uplands is that rainfall and infiltration are the only source/ sink terms when calculating the flow on uplands and the flow condition on uplands is turbulent. Therefore, in order to calculate the flow rate, the Manning’s equation is used:

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 q ¼ y5=3 S0 @y=@x n ð10Þ

W04434

where n is the Manning’s friction coefficient.

4 of 16

ð14Þ


W04434

[22] The upstream surface runoff flowing from urbanized areas (simulated by SWMM5 [Huber and Dickinson, 1988; Rossman, 2005]) is incorporated into the wetland model as an inflow boundary condition at the stream. Overland flow generated from both upland and wetland sites becomes the lateral inflow. It is formulated mathematically by adding the flow generated from the wetland to the stream at each node of discretized stream segments as follows: Qi ¼ Qi þ ðdxÞ ðqwetland Þi¼nn

ð15Þ

[23] The continuity of overland flow between upland and wetland areas is satisfied by using the appropriate boundary conditions at upland/wetland boundaries, which is the continuity of flow rate per unit length:

qupland

i¼nn

¼ ðqwetland Þi¼1 qupland ¼ KS0 ym

"

) B:C: at wetland : yji¼1 ¼

#1=m qupland i¼nn ð16Þ KS0

where i is the node number, nn is the total number of nodes along the upland or wetland area. The flow rate in the stream is calculated by @Q @Q @2Q þc ¼ K1 2 c ¼ mV @t @x @x

K1 ¼

Q 2BS0

ð17Þ

where B is the channel width (L). The flow in the detention area of a wetland site is calculated for a given stagedischarge relationship with a storage routing method as follows: dy I ð y; t Þ Oð y; t Þ ¼ dt Að y; tÞ

ð18Þ

where y is the water depth in lake (L), I is the inflow rate (L3/T), O is the outflow rate (L3/T) and A is the surface area of the lake (L2). The inflow rate includes the rainfall, evapotranspiration and groundwater recharge/discharge in addition to the streamflow rate: I ð y; tÞ ¼ Qs þ Að y; t Þ½qr qet þ qdrch

ð19Þ

[24] Here the exchange between groundwater and the lake is calculated in the vertical direction, as the wetland is interacting with underlying groundwater, and therefore the term qdrch is given by qdrch ¼ Kz

@H @z

0

ð20Þ

for groundwater recharge and discharge, respectively. [25] The average detention time t is then calculated as the time interval between the peak of inflow and outflow rates during a storm event. t ¼ tOmax tImax

ð21Þ

where tImax is the time at which the inflow reaches maximum and tOmax is the time at which the outflow reaches maximum.

W04434

In order to solve the diffusion wave equations that describe overland and channel flow, the MacCormack explicit finite difference method is used for single, discrete storm events, and an implicit finite difference method is used for continuous storm events. Since implicit finite difference methods are unconditionally stable, they provide the advantage of choosing arbitrary time and space steps, particularly advantageous for long-term simulations (as is the case for continuous storm event modeling). On the other hand, the MacCormack method provides more accurate and efficient results for single (discrete) storm event simulations [Kazezyılmaz-Alhan et al., 2005]. In order to solve the storage equation, a fourth-order Runge-Kutta numerical method is used. 2.2. Wetland Water Quality [26] The water quality component of the conceptual wetland model calculates the concentration of nitrogen compounds (organic nitrogen, ammonium nitrogen and nitrate nitrogen) and total phosphorus. There is also the option of calculating the total nitrogen rather than calculating each nitrogen compound in the nitrogen cycle. The one dimensional advection-dispersion-reaction equation is used for each constituent and the equations are coupled for the nitrogen compounds. The vegetation effect of a wetland site is represented by plant uptake/release terms as sources and sinks. The surface water/groundwater interaction effect is taken into account with the terms that represent the incoming/outgoing mass due to groundwater recharge/ discharge. The surface water velocity (in the advection term) is calculated within the water quantity component of the model. The concentration of nitrogen and phosphorus are calculated both on wetland and stream areas and in the lake or storage area. Inputs of nitrogen and phosphorus that are on wetland sites join the stream as lateral inflows. Formulations for phosphorus, nitrogen and nitrogen compounds concentrations are developed by incorporating the reaction sequence for nitrogen and surface water/groundwater interaction terms into the one dimensional advection-dispersion equations as follows: Total phosphorus (TP) @CTP @CTP 1 @ @CTP qLin L ¼ V þ þ CTP CTP Ax Dx @t @x @x Ax Ax @x qgwd gw ðCTP CTP Þ KTP CTP þ Ax ð22Þ

Total nitrogen (TN)

@CTN @CTN 1 @ @CTN qLin L Ax Dx ¼ V þ þ CTN CTN Ax @x @t @x @x Ax qgwd gw CTN CTN KTN CTN þ Ax ð23Þ

Organic nitrogen (ON) @CON @CON 1 @ @CON qLin L Ax Dx CON CON ¼ V þ þ Ax @x @t @x @x Ax qgwd gw CON CON KON CON þ JRON þ Ax ð24Þ

5 of 16

W04434


W04434

Figure 3. (a) Duke University and wetland site watersheds in Durham, North Carolina (35°5901600N, 78°4904500W). (b) Restored wetland and stream site at Sandy Creek in Duke Forest, North Carolina (Duke Wetland Center). Contour lines are shown at 30 cm. The stream and lake restoration areas are shown as darkened areas in the map. Numbers along gray lines (T) indicate water well locations. (c) Discretization of the Duke University wetland site. Ammonium nitrogen (AN)

Nitrate nitrogen (NN)

@CAN @CAN 1 @ @CAN qLin L Ax Dx ¼ V þ þ CAN CAN Ax @x @t @x @x Ax qgwd gw CAN CAN þ KON CON KAN CAN JUAN þ Ax ð25Þ

@CNN @CNN 1 @ @CNN qLin L Ax Dx ¼ V þ þ CNN CNN Ax @x @t @x @x Ax qgwd gw CNN CNN þ yKAN CAN KNN CNN JUNN þ Ax ð26Þ

where C is the concentration (M/L3), CL is the lateral concentration (M/L3), Cgw is the concentration in ground6 of 16

W04434


W04434

Figure 4. Duke University West Campus Drainage System (M.A. Medina Jr., Duke University, The stormwater modeling and management project, personal communication, 2004). The campus is divided into two main watersheds, watershed A and watershed B, shown by the thick line. The dashed thick line shows part of the watershed A from which the storm water drains into the wetland area. The conduits can be channels or pipes.

water (M/L3), K is the first-order loss rate constant (1/T), Ax is the cross-sectional area in x direction (L2), Dx is the dispersion coefficient (L2/T), qLin is the lateral inflow (L2/T), qgwd is the groundwater discharge (L2/T), JRON is the release flux of organic nitrogen from biomass (M/T), JUAN is the uptake flux of ammonium nitrogen absorbed by biomass (M/T), JUNN is the uptake flux of nitrate nitrogen absorbed by biomass (M/T), y is the fraction of ammonium that is nitrified, and TP, TN, ON, AN, NN are the subscripts denoting total phosphorus, total nitrogen, organic nitrogen, ammonium nitrogen, nitrate nitrogen, respectively. [27] In this simplified reaction sequence, the wetland contributes to the amount of organic nitrogen from decomposition of biomass. Then, organic nitrogen is ammonified to ammonium nitrogen. The amount of ammonium nitrogen decreases via volatilization of ammonia, nitrification, sorption, and plant uptake. The fraction of ammonium (y) is nitrified and the remaining fraction (1 y) is gasified. Nitrification forms nitrate nitrogen, and denitrification and uptake decrease nitrate nitrogen. KON represents the reduc-

tion rate of organic nitrogen by ammonification, KAN represents the reduction rate of ammonium by nitrification (yKAN) and volatilization [(1 y) KAN], and KNN represents the reduction rate of nitrate by denitrification. The uptake into living plants, and sorption onto sediment, are taken into account by the uptake flux term. Biomass decomposition is taken into account by the release flux term. Organic nitrogen is released into water because of the biomass decomposition. The total nitrogen calculation includes all the effects (i.e., ammonia volatilization, bacterial immobilization, nitrification, and denitrification) in one term, given by KTN. [28] The main phosphorus compounds in wetlands are dissolved phosphorus, solid mineral phosphorus and solid organic phosphorus [Richardson, 1999]. Soluble phosphorus is taken up by plants or sorbed to wetland soils and sediments. The most important physical processes in the phosphorus removal are sedimentation of particulate phosphorus and sorption of soluble phosphorus. Plant uptake is not a proper measure of phosphorus removal in a wetland

7 of 16

W04434


W04434

Table 1. Parameters of the WETSAND Model as applied to the Duke University Wetland and Stream Restoration Site Variable Landscape characteristics Lw S0 W n

Value

Comment

changes for each section changes for each section changes for each section 0.125 (upland), 0.055 (stream)

K

1 107 m1d1 (dense), 5 107 m1d1 (sparse)

z

2 (new channel), 1.5 (old channel, downstream)

length of each upland, wetland, stream section and lake bed slope of each upland, wetland, stream section width of each upland, wetland, stream section and lake Manning’s friction coefficient; given for Sandy Creek Tributary D (ATC Associates Inc. of North Carolina, PC, Duke wetlands Sandy creek tributary ‘‘D’’ Clomr Fema submittal, 2003) coefficient which reflects vegetation density in the power law formulation [Kadlec and Knight, 1996] side slope of trapezoidal cross section of stream

Surface water/ground water interaction md gwl Kx Infiltration Kz

assumed same as vertical hydraulic conductivity 0.65 103 m h1

qe

0.486

qi

0.45 166.8 103 m

S

days on which groundwater level measurements are taken measured groundwater depth above mean sea level on each wetland section horizontal hydraulic conductivity vertical hydraulic conductivity; given for silt loam type soil [Rawls et al., 1983] effective porosity; given for silt loam type soil [Rawls et al., 1983] initial moisture content; assumed close to qe suction head; given for silt loam type soil [Rawls et al., 1983]

Rainfall qr Evapotranspiration T

rainfall intensity; data collected at Duke Forest site monthly mean temperature; data used from National Climatic Data Center monthly mean day length at given latitude in units of 12 hours; data used for N35 latitude

L Water quality CTPgw

0 mg L1

CTNgw

1.5 mg L1

CONgw CANgw CNNgw KTP

11.5 yr1

KTN

20 yr1

KON KAN KNN JRON JUAN JUNN y Dx

groundwater concentration for total phosphorus; taken from Kadlec and Knight [1996] groundwater concentration for total nitrogen; taken from Kadlec and Knight [1996] groundwater concentration for organic nitrogen groundwater concentration for ammonium nitrogen groundwater concentration for nitrate nitrogen first-order loss rate constant for total phosphorus reduction; taken from Kadlec and Knight [1996] first-order loss rate constant for total nitrogen reduction; taken from Kadlec and Knight [1996] first-order loss rate constant for organic nitrogen reduction first-order loss rate constant for ammonium reduction first-order loss rate constant for nitrate reduction release flux of organic nitrogen from biomass uptake flux of ammonium nitrogen absorbed by biomass uptake flux of nitrate nitrogen abosorbed by biomass fraction of ammonium that is nitrified dispersion coefficient for each substance; taken from Kadlec and Knight [1996]

100 m2 h1 (upland), 10 m2 h1 (wetland), 1000 m2 h1 (stream), 3 m2 h1 (lake)

site because most of the stored phosphorus returns to water by decomposition [Richardson, 1999]. Since overall phosphorus cycling and storage is very complex, quantitative calculations of phosphorus removal are lumped into one term, given by KTP. This term represents the total amount of total phosphorus removed by the wetland. [29] The reaction sequence among organic, ammonium and nitrate nitrogen components is given by Kadlec and

Knight [1996]. In the WETSAND model, this reaction sequence and mass influx/outflux terms (which represent the solute exchange between the surface and subsurface) are incorporated into the one dimensional advection-dispersion equation. A large database would be required in order to be able to conduct a thorough analysis of each nitrogen component individually and most of the time such a database would not be available. Therefore equations (22)

8 of 16

W04434


W04434

Figure 5. Inflow and outflow rates at Sandy Creek during a storm event in August 2002. and (23) could be used for limited data sets which take into account the effect of vegetation uptakes, biomass decay, and nutrient returns with a first-order loss rate constant [Kadlec and Knight, 1996] incorporated into one dimensional advection-dispersion equation. [30] All of these governing equations are solved by an implicit finite difference method. As noted earlier, the concentrations generated from upstream urban areas are simulated using SWMM5 and the results are incorporated into the water quality component of the wetland model as boundary conditions at upstream points of a wetland area. The concentrations generated from wetland areas and flowing into the stream are treated as lateral concentration terms in the stream concentration equations. ðqLin Þstream ¼ qwetland

CL ¼ Cwetland

ð27Þ

[31] The zero flux boundary condition is employed as a downstream boundary condition. @C ¼0 @x x¼L

ð28Þ

3. Model Application [32] The WETSAND model has been applied to the Duke University restored wetland site and the predicted results of hydrologic conditions and water quality are discussed in the following sections.

3.1. Study Site [33] The Duke University restored wetland site is located in the Sandy Creek watershed, which is in the southern section of Durham County and covers 554.41 ha (1,370 acres). Storm water runoff flows into the wetland area from part of the Duke University campus and part of the City of Durham, NC (Figure 3a). Sandy Creek is a tributary of the New Hope Creek and New Hope Creek flows into the Jordan Reservoir, which feeds into the Cape Fear River. The Jordan Reservoir and Cape Fear River are essentially the drinking water supplies of a large portion of the state of North Carolina, including Durham and Orange counties. The wetland stream restoration project is designed to transform the degraded portion of Sandy Creek into 2 ha (5 acres) of wetland by recontouring and raising the water table of the creek. Over 579 m (1900 ft) of Sandy Creek have been restored by closing part of the original streambed and opening a new streambed with more meanders, in order to enhance water flow over the floodplain and aid in the removal of nutrients and sediments. [34] Sandy Creek has many tributaries within Duke University boundaries, through which the storm water flows into the wetland area (Figure 3a). Figure 3b shows the topography of the restored wetland site, and new restored stream and well locations. The wetland along the stream was restored in the fall of 2004. There are 20 groundwater sampling wells located on 11 transects (gray lines (T01 – T11) in Figure 3b) throughout the wetland site and the groundwater level measurements are taken once every two weeks. Groundwater samples were collected two years prior to construction to determine shifts in water table due to

9 of 16

W04434


W04434

Figure 6. Inflow to and outflow from the lake at the wetland site from continuous simulation for year 2002. restoration. The slotted wells are 120 cm in depth. The construction of an earthen dam was completed in October of 2005 and the flooded area behind the dam is shown on the map (Figure 3b). The flooded area is a lake, which will normally have a surface water level at 89.92 m (295 ft) above mean sea level. The lake has a weir gate which will allow for the raising and lowering of the lake 1 m so that water levels in the stream and wetlands can be altered. Water quality measurements for major nutrients (N, P and cations) are taken monthly at the wetland and lake site as well as in all tributaries [Richardson et al., 2006]. Additional measurements of nutrients are taken at six locations (S1– 6, Figure 3c) within the restored stream as well as in the wetland, if standing water is present. 3.2. Study Results [35] The WETSAND model has been applied to the Duke University restored wetland site to predict if decreases in the peaks of surface flows and improved water quality can be expected as a result of the wetland and stream restoration. Figure 3c shows the discretization of the Duke University wetland site into six upland, ten wetland and six stream segments. The boundary between an upland and wetland section was selected according to the land slope change, such that uplands were on the steep slope side and wetlands were on the mild slope side. The transition from upland to wetland occurs along a boundary that is approximately at 91.14 m (299 ft) elevation. Figure 4 shows the drainage system of Duke University West Campus that is simulated using SWMM5. The campus is divided into watershed A and watershed B (thick black line in Figure 4). Storm water from part of watershed A (shown with thick dashed black

line) and from all of watershed B drains into the wetland area through the nodes denoted as 335 and 329, respectively. The calculated outflow rates at node 335 in watershed A and at node 329 in watershed B are routed into the wetland area through Sandy Creek Tributary D as shown in Figure 3c. The overland flow generated from both university campus and wetland sections is collected through the Sandy Creek channel into the lake area, the latter decreases the peak flow rates and increases the hydraulic residence time of water in the wetland area to improve water quality. [36] For Sandy Creek Tributary D, the Manning coefficient is given in the range of 0.08 – 0.17 for overbank sections and 0.050 – 0.059 for channel sections (ATC Associates Inc. of North Carolina, PC, Duke wetlands Sandy creek tributary ‘‘D’’ Clomr Fema submittal, 2003). In the simulations, the Manning coefficient is selected as 0.125 and 0.055 for upland sections and the stream channel, respectively. From analyses conducted for sparse vegetation types, a coefficient K equal to 5 107 was selected to calculate overland flow over the wetland sections. The soil type is silt loam. Currently, a detailed aquifer characterization is not available at the site and therefore an analysis for anisotropic transmissivity is not possible. For this reason, we assume the lateral hydraulic conductivity to be equal to the vertical hydraulic conductivity of silt loam soil for our analysis, although WETSAND can simulate the results for soils with anisotropic hydraulic properties. The stream has a trapezoidal cross-section at each point. The slope of the newly restored channel is 0.0018 with side slopes of 2 and a bottom width of 4.4 m. The tributary beginning with node 335 has a slope of 0.0182 with side slopes of 2 and bottom

10 of 16

W04434


W04434

Figure 7. Change of water depths and groundwater contours on wetland section 4 and Wetland section 8 during August 2002. width of 4.57 m. The downstream portion (which flows into the lake area) has a slope of 0.0011, with side slopes of 1.5 and bottom width of 6.7 m. The input parameters that are used in the WETSAND model for the Duke University restored wetland site are summarized in Table 1. [37] Simulations of the wetland area are conducted for the year 2002, using rainfall data collected at the nearby Duke Forest Site. The rainfall depth (mm) is recorded at every half hour throughout the year by the FACE (Free Air Carbon Dioxide Enrichment) Facility at Duke University. This facility occupies about 23 hectares (57 acres) of pine forest within the Blackwood division and has structures and utilities associated with the Forest Atmosphere Carbon Transfer and Storage (FACTS-1) experiment. Rainfall data are collected in addition to many other meteorological and

environmental data. Rainfall intensity is determined from the recorded rainfall depth. Groundwater level data were collected by Duke University Wetland Center investigators every 14 days from 20 sampling wells located throughout the wetland site: these data are used in the simulation of surface water/groundwater interaction effects. The average monthly temperature for Durham, NC area was obtained from National Climatic Data Center sources [EarthInfo Inc., 2004] and was used in the Thornthwaite evapotranspiration estimates in the wetland model. [38] In order to test the WETSAND model performance and capability, the inflow at upstream points of Sandy Creek Tributary D is routed through the channel and the outflow at downstream points of the Sandy Creek is calculated using WETSAND. The inflow is the storm water which drains

11 of 16

W04434


W04434

Figure 8. Groundwater contours above mean sea level on 16 August 2002. The well locations are shown along with the measured water levels on the plot. from part of watershed A and from all of watershed B through node 329 into the wetland area, obtained using SWMM5. Figure 5 shows the inflow and outflow rates at these upstream and downstream points, along with the rainfall intensity time series during an August 2002 storm event. As seen from Figure 5, the peaks of the hydrographs follow the peaks of the rainfall with outflow rates lagging the inflow rates due to roughness, storage, diffusion effects and other surface mechanics effects. Therefore the WETSAND model predicts the outflow reasonably well. The outflow from the lake is calculated by using equations (18) and (19) to predict the lake detention effect on peak flows during storm events. Figure 6 shows the continuous simulation results for year 2002. It can be observed that the peak flows decrease significantly due to lake detention effects. The average detention time of storm water in the lake is calculated by using equation (21) and it is calculated as 5 hours. As a result, the restored wetland/stream/lake site attenuates the peak of flow rate during storm events. [39] The change in water depth on wetland sections four (W4) and eight (W8) are investigated in detail because a groundwater recharge/discharge effect is observed at these two specific land sites. The length and width of wetland sections four (W4) and eight (W8) are approximately 78 93 m and 29 83 m, respectively. Figure 7 shows the change in water depth on wetland sections four (W4) and eight (W8) for the cases for which surface water/

groundwater interaction effects are included, and for which no interaction effect is simulated for a storm event during August 2002. The groundwater contours are plotted using the measured data set obtained from wells located at the wetland site on 16 August 2002 (Figure 7). The groundwater contours are shown in detail in Figure 8, along with the well locations and the measured water levels at these wells. On wetland section eight (W8), with the interaction effect included, the water stays on the wetland surface longer than for the case with no interaction effect included (Figure 7). This behavior can be confirmed by noting the groundwater level contours, which suggest a groundwater discharge on this particular wetland section. On wetland section four (W4), with the interaction effect included, a faster decrease in the surface water depth is observed. Again, this behavior can be confirmed by noting the groundwater level contours, which suggest a groundwater recharge on this particular wetland section. [40] For the water quality simulations of the wetland site, measured total nitrogen and total phosphorus data are used, collected by Duke University Wetland Center investigators between August – October 2002 in the Sandy Creek Tributary D at the wetland site. The water quality data was collected for every hour for several days (15 August, 18 August, 23 August, 26 August, 28 August, 30 August, 14 September, 26 September, and 15 October) for different time periods. Hydraulic routing through Sandy Creek Trib-

12 of 16

W04434


W04434

Figure 9. Inflow concentrations from Sandy Creek into the lake, outflow concentrations from the lake, and total nitrogen data.

Figure 10. Inflow concentrations from Sandy Creek into the lake, outflow concentrations from the lake, and total phosphorus data. 13 of 16

W04434


utary D and through the lake was conducted in order to predict the effect of lake detention on water quality. [41] For the water quality analyses, the dispersion coefficients for upland sections, wetland sections, streams and lake are assumed as 100, 10, 1000, and 3 m2/hr, respectively [Kadlec and Knight, 1996]. The groundwater concentrations for total phosphorus and total nitrogen are assumed to be 0 mg/L and 1.5 mg/L, respectively [Kadlec and Knight, 1996]. The first-order loss rate constants are assumed to be 11.5 yr1 and 20 yr1 for total phosphorus and total nitrogen, respectively [Kadlec and Knight, 1996]. [42] Figure 9 shows the continuous simulation results for the prediction of total nitrogen reduction during August – October 2002. The measured concentration levels are used as boundary conditions upstream of Sandy Creek. The concentration levels at a downstream point of Sandy Creek and at the outlet of the lake are calculated by using the WETSAND model. Only the maximum values of the total nitrogen concentration data for each measurement set taken on different dates are shown in comparison to predicted peak concentrations to avoid clutter. The actual measured nitrogen concentrations collected on 15– 16 October 2002 at an upstream point of Sandy Creek and the concentrations calculated at a downstream point of Sandy Creek by using WETSAND are shown in greater detail in the inset. As noted from Figure 9, the WETSAND model predicts that the peaks of total nitrogen have a minor decrease when the flow reaches downstream segments of Sandy Creek. This shows that we do not achieve significant improvement in the water quality with base flow in the wetland/stream site. However, after the flow reaches the ponded lake area, a significant decrease in the peak concentrations is predicted. This is because of the detention effect within the lake area at the downstream section of the wetland site. As a result, we predict a significant improvement in water quality at the wetland/stream/storm water lake complex site before the water flows to water reservoirs located further downstream. [43] Figure 10 shows the continuous simulation results for the prediction of total phosphorus reduction during August – October 2002. We use the measured total phosphorus concentration data as the boundary condition at an upstream point of Sandy Creek and present the simulation results of WETSAND at a downstream point of Sandy Creek (i.e., at the inlet of the lake and at the outlet of the lake). The measured data set for total phosphorus is routed through Sandy Creek and through the lake area. Only the maximum values of the total phosphorus concentration data for each measurement set collected on different dates are shown along with the peak concentrations calculated using WETSAND. The actual measured phosphorus concentrations collected on 15– 16 October 2002 at an upstream point of Sandy Creek and the predicted concentrations calculated at a downstream point of Sandy Creek are shown in greater detail in the inset. The simulation results for the prediction of total phosphorus concentration levels also show a significant decrease in the peak concentrations when the storm water has been routed through the lake. The peak concentration does not decrease significantly downstream of Sandy Creek before it reaches the ponded area of the restored wetland/stream site. The water quality prediction shows that the entire restored wetland site complex has an important impact on the water supply by decreasing the concentration

W04434

levels in the water before it flows further downstream. However, base flow nutrients are not significantly reduced within the stream alone, especially when contact with the adjacent wetlands is minimal under lowest flow conditions [Richardson et al., 2006].

4. Discussion [44] WETSAND is a comprehensive model, which combines a large number of processes, including coupled surface water/groundwater components for nutrient transport and hydrological processes for wetland sites. The WETSAND model takes into account the flow from groundwater to surface water or from surface water to groundwater as a direct effect on water levels at a wetland site. Consequently, the different fluctuations observed in water levels due to the surface water/groundwater interaction indirectly affect the nutrient transport. Furthermore, WETSAND simulates the direct effect of water flux between the surface and subsurface phases on the nutrient concentration, with mass influx and mass outflux terms included in the advection-dispersion-reaction equations for nitrogen and phosphorus. Therefore the surface water/ groundwater interaction is combined with nutrient transport. [45] The results for the Duke University restored wetland/ stream site are presented to demonstrate the model capabilities and its contributions to wetland hydrology and quality. There are some uncertainties in the results, due to assumptions made in model parameters, such as first-order loss rate constants for nitrogen and phosphorus, and dispersion coefficients for wetland, stream and lake areas. Extensive data are still needed to be able to calibrate the model further and verify the simulation results. Current studies are now underway at the site to collect this data now that the storm water lake is in place and the system is fully operational. Nevertheless, the outcomes of this study are important, as they show the significance of surface water/groundwater interaction on hydrological processes and contaminant transport at a wetland site. Furthermore, WETSAND can be used to model other riverine or constructed wetland sites. [46] The total nitrogen and phosphorus removal in wetlands is an integrative measure of individual transformations. The first-order loss rate constant K for total nitrogen and total phosphorus is the average value of the constant K for different wetland sites which are obtained by fitting the plug flow model to measured data at these different sites [Kadlec and Knight, 1996]. The data collected at wetlands reflect ammonia volatilization, bacterial immobilization and nitrification and denitrification for total nitrogen. The factors that affect the first-order loss rate constant are loading rate, climate, plant community composition and soil characteristics. The range of the value of dispersion coefficients for wetlands, ponds and rivers is discussed by Kadlec and Knight [1996]. The values are given for different sites and an approximate number is assumed based on these values for the analyses. The dispersion coefficients should be site specific and will be a future study for the Duke University wetland site. [47] The surface water/groundwater interactions are commonly calculated with a simple Darcy’s equation or with the modified versions of Green-Ampt infiltration equation or with Richard’s equation [Langevin et al., 2004]. The mathematical representation of surface water/groundwater inter-

14 of 16

W04434


actions may be very complicated; therefore further research is required to describe and represent the actual physical mechanisms mathematically. Sophocleous [2002] discusses the complications of surface water/groundwater interactions and the need for modifying Darcy’s law for different hydrological conditions. In our model, we use the simple Darcy’s equation for wetlands where the unsaturated layer is thin and groundwater levels are close to the wetland elevations. The Duke University restored wetland site is such an example. The WETSAND model calculates the surface water/groundwater interactions at the banks of the wetland with the Darcy’s equation in the horizontal direction and at the lake of the wetland with the Darcy’s equation in the vertical direction. [48] MODFLOW, originally developed for groundwater flow modeling, has recently been used by some investigators to simulate the wetland subsurface flow. These investigators either utilize MODFLOW directly to simulate wetland hydrology [Bradley, 1996, 2002] or have developed additional packages to MODFLOW [Restrepo et al., 1998; South Florida Water Management District, Wetland Simulation Model, http://www.sfwmd.gov/org/pld/hsm/modflow/ modflow.htm]. These models have the advantages of simulating the subsurface flow and the wetland-aquifer interaction in wetlands (better for subsurface flow wetland types). However, they are not adequate to simulate the wetland hydrology for surface flow wetland types. The WETLAND model [Lee et al., 2002] and the Dynamic Model for Storm water Treatment Areas (DMSTA) (http:// www.walker.net/dmsta/index2.htm) simulate wetland hydrology by using a lumped parameter system. The advantage of these models is that they require fewer input parameters; on the other hand, they don’t provide much information about the differences in the hydrologic regime at particular sections within the wetland site. Furthermore, these models do not have surface water/groundwater interaction components. [49] The WETSAND model can be developed further by incorporating a subsurface flow component for the vadose zone and a groundwater flow component for the saturated zone. Further development will include incorporating a nutrient/sediment transport module, and investigating the fluid dynamics and biogeochemistry nonlinearities which will amplify projected changes in the hydrologic regime in seasonal wetlands. The nutrient/sediment transport module can be extended by including total suspended sediments, and the carbon cycle and dissolved oxygen tied in nitrogen and phosphorus dynamics. Currently, WETSAND is a one dimensional model which is adequate to model elongated wetland sites such as the one in the Sandy Creek watershed. However, the model can be extended into two or three dimensions so that it can handle the wetland hydrological processes with any type of geometry.

5. Conclusions [50] A comprehensive conceptual wetland model Wetland Solute Transport Dynamics (WETSAND) has been developed that predicts both water quantity and water quality of a wetland site. The effect of surface water/groundwater interactions on wetland hydrology and water quality has been investigated by incorporating groundwater recharge/ discharge terms into the wetland model. The WETSAND

W04434

simulation results show that the effects of surface water/ groundwater interactions in general play a significant role on wetland dynamics and therefore should be taken into account when modeling wetland hydrology and solute transport. The output of the EPA Storm Water Management Model (SWMM5) is incorporated into the WETSAND model to compute the surface runoff and water quality components flowing into a wetland area from upstream urbanized areas. The water quantity component of the WETSAND model accounts for rainfall, evapotranspiration, infiltration and groundwater recharge/discharge as water source/sink terms. The water quality component of the WETSAND model calculates the concentration of nitrogen compounds or total nitrogen and total phosphorus. [51] The Duke University restored wetland/stream site was selected as the application site of the WETSAND model. After the WETSAND model was applied to this restored wetland/stream/lake complex site, it was observed that while groundwater recharges, water depths in wetland areas drop faster during storm events than for the case with no groundwater recharge. When groundwater discharges, water depth decays more gradually than for the case with no groundwater discharge. Therefore it is concluded that surface water/ groundwater interactions affect surface water depths, and therefore the hydraulic regime of the wetland site. [52] Analysis of peak flows during storm events shows that they decrease significantly due to storm water lake detention effects. The water quality model results indicate that the peaks of contaminant concentrations decrease significantly after detention through the ponded areas on the wetland site. Thus restoring the wetland site is important and can be used as a best management practice for peak flow control during storm events, and for water quality improvement. [53] The WETSAND model can be used to simulate the fluctuations in the amount of water and in the magnitude of pollutants at different wetland sites through which water flows into the receiving water. The effect of hydroclimatic variability, such as temperature and rainfall, on streamflow and groundwater inputs in wetlands, and therefore its impact on nitrogen and phosphorus concentrations, can also be investigated by using the WETSAND model. Both deterministic and probabilistic analyses can be conducted to simulate the changes in wetland dynamics under different rainfall regimes. [54] Acknowledgments. The authors would like to express their gratitude to the anonymous reviewers and the Associate Editor, Rob Runkel, for their excellent suggestions, which strengthened the paper. Partial funding for this investigation was provided by the Facilities Management Department of Duke University, and the authors wish to acknowledge in particular Glenn Reynolds and Gary Teater (Systems/Engineering Services) and Raymond Wrenn (Construction Services) for valuable discussions on storm water and wetland hydrologic and water quality modeling needs on campus. Rainfall data were obtained from a project supported by the Office of Science (BER), U.S. Department of Energy, grant DE-FG02-95ER62083. The authors wish to acknowledge also the Duke Wetland Center Case Studies Program for funding water quality studies and the NC Clean Water Management Trust Fund for funding water level data collections and, most importantly, funding the restoration of the stream/wetland lake complex in the Duke Forest.

References Borin, M., G. Bonaiti, and L. Giardini (2001), A constructed surface flow wetland for treating agricultural waste waters, Water Sci. Technol., 44(11 – 12), 523 – 530.

15 of 16

W04434


Bradley, C. (1996), Transient modeling of water-table variation in a floodplain wetland, Narborough Bog, Leicestershire, J. Hydrol., 185, 87 – 114. Bradley, C. (2002), Simulation of annual water table dynamics of a floodplain wetland, Narborough Bog, UK, J. Hydrol., 261, 150 – 172. Choi, J., and J. W. Harvey (2000), Quantifying time-varying ground-water discharge and recharge in wetlands of the northern Florida Everglades, Wetlands, 20(3), 500 – 511. Chu, S. T. (1978), Infiltration during an unsteady rain, Water Resour. Res., 14(3), 461 – 466. Devito, K. J., and A. R. Hill (1997), Sulphate dynamics in relation to groundwater-surface water interactions in headwater wetlands of the southern Canadian Shield, Hydrol. Processes, 11(5), 485 – 500. Dorge, J. (1994), Modelling nitrogen transformations in freshwater wetlands: Estimating nitrogen retention and removal in natural wetlands in relation to their hydrology and nutrient loadings, Ecol. Modell., 75/76, 409 – 420. EarthInfo Inc. (2004), USGS daily values east [CD-ROM], report, Boulder, Colo. Green, W. H., and G. A. Ampt (1911), Studies on soil physics, 1, The flow of air and water through soils, J. Agric. Sci., 4(1), 1 – 24. Hammer, D. E., and R. H. Kadlec (1986), A model for wetland surface water dynamics, Water Resour. Res., 22(13), 1951 – 1958. Huber, W. C. and R. E. Dickinson (1988), Storm Water Management Model, version 4, user’s manual, Environ. Res. Lab., Athens, Ga. Kadlec, R. H. (1990), Overland flow in wetlands: Vegetation resistance, J. Hydraul. Eng., 116(5), 691 – 706. Kadlec, R. H., and D. E. Hammer (1988), Modeling nutrient behavior in wetlands, Ecol. Modell., 40, 37 – 66. Kadlec, R. H. and R. L. Knight (1996), Treatment Wetlands, CRC Press, Boca Raton, Fla. Kadlec, R. H., D. E. Hammer, I. Nam, and J. O. Wilkes (1981), The hydrology of overland flow in wetlands, Chem. Eng. Commun., 9, 331 – 344. Kazezylmaz-Alhan, C. M., M. A. Medina, Jr., and R. Prasada (2005), On numerical modeling of overland flow, Appl. Math. Comput., 166, 724 – 740. Keefe, S. H., L. B. Barber, R. L. Runkel, J. N. Ryan, D. M. McKnight, and R. D. Wass (2004), Conservative and reactive solute transport in constructed wetlands, Water Resour. Res., 40, W01201, doi:10.1029/ 2003WR002130. Krasnostein, A. L., and C. E. Oldham (2004), Predicting wetland water storage, Water Resour. Res., 40, W10203, doi:10.1029/2003WR002899. Langevin, C. D., E. D. Swain, and M. A. Wolfert (2004), Simulation of integrated surface-water/ground-water flow and salinity for a coastal wetland and adjacent estuary, U.S. Geol. Surv. Open File Rep., 2004-1097, 30 pp. Lee, E. R., S. Mostaghimi, and T. M. Wynn (2002), A model to enhance wetland design and optimize nonpoint source pollution control, J. Am. Water Resour. Assoc., 38(1), 17 – 32. Martin, J. F., and K. R. Reddy (1997), Interaction and spatial distribution of wetland nitrogen processes, Ecol. Modell., 105, 1 – 21. McKillop, R., N. Kouwen, and E. D. Soulis (1999), Modeling the rainfallrunoff response of a headwater wetland, Water Resour. Res., 35(4), 1165 – 1177. Mein, R. G., and C. L. Larson (1973), Modeling infiltration during a steady rain, Water Resour. Res., 9(2), 384 – 394.

W04434

Mitchell, G. F., C. L. Hunt, and Y. M. Su (2002), Mitigating highway runoff constituents via a wetland, Transp. Res. Rec., 1808, 127 – 133. Mitsch, W. J. and J. G. Gosselink (2000), Wetlands, 3rd ed., John Wiley, Hoboken, N. J. Moore, M. T., R. Schulz, C. M. Cooper, and J. H. Rodgers (2002), Mitigation of chlorpyrifos runoff using constructed wetlands, Chemosphere, 46(6), 827 – 835. Price, J. S., and J. M. Wadington (2000), Advances in Canadian wetland hydrology and biochemistry, Hydrol. Processes, 14(9), 1579 – 1589. Rawls, W. J., D. L. Brakensiek, and N. Miller (1983), Green-Ampt infiltration parameters from soils data, J. Hydraul. Eng., 109(1), 62 – 70. Restrepo, J. I., A. M. Montoya, and J. Obeysekera (1998), A wetland simulation module for the MODFLOW ground water model, Ground Water, 36(5), 764 – 770. Richardson, C. J. 1995), Wetlands ecology, in Encyclopedia of Environmental Biology, vol. 3, pp. 535 – 550, Elsevier, New York. Richardson, C. J. (1999), The role of wetlands in storage, release, and cycling of phosphorus on the landscape: A 25 year retrospective, in Phosphorus Biogeochemistry in Sub-tropical Ecosystems, edited by K. R. Reddy, pp. 47 – 68, CRC Press, Boca Raton. Richardson, C. J., S. Qian, C. B. Craft, and R. G. Qualls (1997), Predictive models for phosphorus retention in wetlands, Wetlands Ecol. Manage., 4, 159 – 175. Richardson, C. J., R. Elting, J. Pahl, and G. Katul (2006), Restoration of hydrologic and biogeochemical functions in bottomland hardwoods, in Hydrology and Management of Forested Wetlands: Proceedings of the International Conference, pp. 50 – 57, Am. Soc. of Agric. and Biol. Eng., St. Joseph, Mich. Rossman, L. A. (2005), Storm Water Management Model, user’s manual, version 5, Rep. EPA/600/R-05/040, Natl. Risk Manage. Res. Lab., Cincinnati, Ohio. Sophocleous, M. (2002), Interactions between groundwater and surface water: The state of the science, Hydrogeol J., 10, 52 – 67. Thornthwaite, C. W. (1948), An approach toward a rational classification of climate, Am. Geogr. Rev., 38, 55 – 94. Warren, F. J., J. M. Waddington, R. A. Bourbonniere, and S. M. Day (2001), Effect of drought on hydrology and sulphate dynamics in a temperate swamp, Hydrol. Processes, 15(16), 3133 – 3150. Winter, T. C. (1999), Relation of streams, lakes, and wetlands to groundwater flow systems, Hydrogeol. J., 7, 28 – 45. Winter, T. C., and D. O. Rosenberry (1995), The interaction of ground water with prairie pothole wetlands in the Cottonwood Lake area, eastcentral North Dakota, 1979 – 1990, Wetlands, 15(3), 193 – 211. Winter, T. C., J. W. Harvey, O. L. Franke, and W. M. Alley (1998), Ground water and surface water: A single resource, U.S. Geol. Surv. Circ., 1139.

C. M. Kazezyılmaz-Alhan, Department of Civil Engineering, Auburn University, Auburn, AL 36849-5337, USA. M. A. Medina Jr., Department of Civil and Environmental Engineering, Box 90287, Duke University, Durham, NC 27708-0287, USA. (miguel. [email protected]) C. J. Richardson, Duke University Wetland Center, Nicholas School of the Environment and Earth Sciences, Duke University, Durham, NC 277080287, USA.

16 of 16