Development of a constructed subsurface-flow wetland simulation model

Ecological Engineering 16 (2001) 519 – 536

www.elsevier.com/locate/ecoleng

Development of a constructed subsurface-flow wetland simulation model Theresa Maria Wynn a,*, Sarah K. Liehr b,1 a

Biological Systems Engineering, Virginia Tech 200 Seitz Hall (0303), Virginia Tech, Blacksburg, VA 24061 -0303, USA b Biological and Agricultural Engineering, P.O. Box 7625, NCSU, Raleigh, NC 27695 -7625, USA Received 16 December 1998; received in revised form 1 May 2000; accepted 10 June 2000

Abstract This paper presents a mechanistic, compartmental simulation model of subsurface-flow constructed wetlands. The model consists of six submodels, including the nitrogen and carbon cycles, both autotrophic and heterotrophic bacteria growth and metabolism, and water and oxygen balances. Data from an existing constructed wetland in Maryland were used to calibrate the model. Model results reproduced seasonal trends well. Interactions between the carbon, nitrogen, and oxygen cycles were evident in model output. In general, effluent biochemical oxygen demand, organic nitrogen, ammonium and nitrate concentrations were predicted well. Because little is known about rootzone aeration by wetland plants, oxygen predictions were fair. The model is generally insensitive to changes in individual parameters. This is due to the complexity of the ecosystem and the model, as well as the numerous feedback mechanisms. The model is most sensitive to changes in parameters that affect microbial growth and substrate use directly. This dynamic, compartmental, simulation model is an effective tool for evaluating the performance of subsurface-flow constructed wetlands. The model provided insights into treatment problems at an existing constructed wetland. With further evaluation and refinement, the model will be a useful design tool for subsurface-flow constructed wetlands. © 2001 Elsevier Science B.V. All rights reserved. Keywords: Constructed wetlands; Monod; Wastewater treatment; CSTR; Subsurface-flow wetland; Nitrogen removal; BOD; Simulation model

1. Introduction Past research has shown constructed wetlands efficiently remove nitrogen and constituents that * Corresponding author. Tel.: +1-540-2316509; fax: +1540-2313199. E-mail addresses: [email protected] (T.M. Wynn), [email protected] (S.K. Liehr). 1 Tel.: + 1-919-5156761; fax: + 1-919-5157760.

create biochemical oxygen demand (BOD) from municipal and industrial wastewaters (Hammer, 1989; Conley et al., 1991; Knight, 1994). While wetland systems are not well understood and few good design data exist, many of the individual processes have been studied in detail (Heliotis and DeWitt, 1983). Treatment processes in constructed wetlands include sedimentation and filtration, precipitation, sorption, and microbial

0925-8574/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 5 - 8 5 7 4 ( 0 0 ) 0 0 1 1 5 - 4

520

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

decomposition, nitrification, and denitrification. Unlike solids and phosphorus removal, nitrogen and BOD removal at low concentrations are mostly the result of complex, interdependent, microbial transformations. Much of the treatment in wetlands is due to both heterotrophic and autotrophic microbes (Kadlec, 1986a; Mitsch and Jorgensen, 1989). Decomposition and ammonification rates are linked to microbial energy requirements, the carbon to nitrogen ratio (C/N) of the organic matter, and the growth rate of microbes (Parnas, 1975; Fyock, 1977; Patrick Jr, 1982; Reddy and D’Angelo, 1997). High removal rates can be sustained in constructed wetlands, although these reactions should be modeled concurrently to better understand and optimize the internal mechanisms (Howard-Williams, 1985; Kadlec, 1986b). In current wastewater wetland design models, the importance of microbial growth on nitrogen and carbon cycling is not explicitly addressed (Reed, 1990; Bavor et al., 1991; Watson et al., 1991; Reed and Brown, 1992; EPA, 1993; Reed et al., 1995). Reactions are lumped into zero- or first-order black-box kinetic models and plug flow hydraulics are used (Reddy and Patrick, 1983; Kadlec, 1994; Kadlec and Knight, 1996). Rate constants are adjusted only for changes in pH and temperature. By assuming transformations, such as nitrification and denitrification, are dependent only on initial concentration (N or BOD), with little connection to microbial metabolism and growth, these models often produce inadequate descriptions of pollutant removal (King et al., 1997). Other, more complex, models also do not model microbial dynamics explicitly, require extensive input data, and are not practical for engineering applications (Bender, 1976; Costanza and Sklar, 1985; Chescheir et al., 1987; Brown, 1988; Jorgensen et al., 1988; Kadlec and Hammer, 1988; Buchberger and Shaw, 1995). A model is needed that describes the individual processes within a constructed wetland system, but that uses readily available site data. The objectives of this research were to develop a mechanistic, compartmental simulation model of constructed subsurface-flow wetlands, to incorporate microbial dynamics in the model, to evaluate constructed wetland system

dynamics, and to provide a tool for constructed wetland design. For simplicity, this research focused only on those processes relevant to wastewater treatment, including N and BOD removal. Monod kinetics were utilized to describe microbial growth rate as a function of substrate(s) availability and the requirements of the specific organism(s). Transformations, such as nitrification and denitrification, were then linked directly to microbial growth. Phosphorus and TSS removal were not modeled because they are relatively well understood, physical processes in constructed wetlands that are generally not dependent on microbial interactions. Experience with both natural and constructed wetlands shows that, eventually, the cation exchange capacity (CEC) will become saturated and only soil accretion and the release of microbial gases will be permanent nutrient sinks (Kadlec, 1986a,b; Brix, 1994). To test the model, data from an existing constructed wetland were compared to model results. Model input parameters were estimated from domestic wastewater and wetland literature, while climate data were obtained from NOAA (National Oceanographic and Atmospheric Administration) records. A sensitivity analysis was conducted to identify those parameters that have the most impact on model predictions.

2. Model development A dynamic compartmental simulation model was developed to model nutrient dynamics in constructed subsurface-flow wetlands. The model consists of six linked submodels representing the carbon cycle, the nitrogen cycle, an oxygen balance, autotrophic bacteria growth, heterotrophic bacteria growth, and a water budget. The wetland is assumed to act as either a single continuously stirred tank reactor (CSTR) or a series of CSTRs. While many wetlands are simulated as plug flow reactors, actual flow conditions are between plug flow and completely mixed, and a CSTR model may be an alternative (Kadlec, 1994; Reed, 1995). Hammer (1984) successfully modeled nutrient dynamics in the Houghton peat-

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

land assuming a well-mixed cell. Additionally, series of CSTRs can be used to simulate nonideal plug flow and lend themselves to compartmental models. The model was developed using simulation software for Apple Macintosh called STELLA II (High Performance Systems, 1992). Due to the graphical nature of STELLA, entire variable names are used instead of symbols. For this model, state variables are in all capital letters, flows are in all lowercase letters, and all other parameters have only the first letter capitalized.

2.1. Carbon and nitrogen cycles The carbon cycle consists of five state variables: BIOMASS (plants), STANDING DEAD, particulate organic carbon (POC), dissolved organic carbon (DOC), and REFRACTORY C (Appendix A). BIOMASS represents wetland plants and associated periphyton. STANDING DEAD refers to the dead above-ground plant parts before they fall to the ground and become plant litter. POC is defined as a homogeneous mixture of decaying plant litter, sloughed microbial cells, and particulate influent wastewater BOD (Jansson and Persson, 1982). DOC is a homogeneous mixture of dissolved influent BOD and carbon compounds leached from decaying plant material. POC and DOC are assumed to have the same chemical composition as domestic wastewater (C10H19O3N), which is also a homogeneous mixture of organics (McCarty, 1975). The processes modeled in the nitrogen cycle include ammonification, immobilization, nitrification, denitrification, and peat accumulation (Appendix A). The state variables are dissolved organic nitrogen (DON), particulate organic nitrogen (PON), ammonia and ammonium nitrogen (NH4), nitrate nitrogen (NO3), IMMOBILIZED N, and REFRACTORY N. Ammonium and ammonia are referred to as ammonium only, since at near neutral pH, ammonium is the predominant form. Atmospheric deposition and ammonia volatilization are not considered. Sorption of NH+ 4 to the soil and organic matter is also not modeled because a simple dynamic sorption

521

model is not available and because sorbed ammonium is still available to attached microbes. Carbon and nitrogen enter the wetland as influent BOD (DOC and POC) and influent nitrogen, and from biomass and microbe death and decay. Carbon is utilized by heterotroph microbes as an energy source and is lost with BOD outflow (DOC only) and peat accumulation. Nitrogen is utilized by autotrophic microbes as an energy source and by microbes and plants as a nutrient. Nitrogen is lost with effluent wastewater, denitrification, and peat accumulation. Microbes are not considered part of POC or PON until they die. For simplification, it is assumed that microbial death contributes only to POC and PON, not DOC or DON, since most microbes in wetlands are associated with plant litter and soil organic matter. Removal of particulate constituents in the wetland, such as particulate organic carbon and particulate organic nitrogen, is assumed 100%. This simplification is necessary because there are no mechanistic models for particulate removal in porous substrates. Since constructed wetland effluent TSS values are consistently low, little error should be introduced by this assumption (Brix, 1994). Plant growth is affected by many factors including solar radiation, temperature, water, and available nutrients. Since solar radiation data are not readily available, a very simple plant model is used, similar to the one described previously by Hammer (1984). Considering nutrients and water are likely not limiting, plant growth is assumed to take place at a constant rate during the growing season. Brown (1988) modeled nutrient uptake as the product of biomass growth and biomass nutrient concentration. Nitrogen uptake is calculated as the product of the C/N of the wetland biomass and biomass growth. Following the end of the growing season, an initial 15% of the plant carbon and nitrogen are lost rapidly due to leaching and physical degradation (Kulshreshtha and Gopal, 1982; Polunin, 1982; Heliotis and DeWitt, 1983; Kadlec, 1986a). The remainder degrades over 1 year to become particulate carbon and nitrogen (Kadlec, 1989b; Johnston, 1991). Below-ground

522

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

biomass is not modeled because roots and rhizomes have a lifespan of 2 – 3 years and little is known about the dynamics of below-ground parts. The model of carbon and nitrogen immobilization, mineralization and ammonification is based on the work of Parnas (1975). Due to the presence of anaerobic microsites at high DO concentrations and aerobic rootzones at low DO concentrations, both anaerobic and aerobic bacteria are always present. The relative fraction of each and the average heterotroph yield are assumed to be a function of DO concentration, reflecting an increasing dominance of energy-efficient aerobes at higher DO concentrations. Autotroph and heterotroph growth are determined during each time step, based on Monod kinetics. The C/N of the wetland determines if the availability of carbon or nitrogen controls the mass transfers of nutrients by heterotrophs. Above 5 mg/l DO, the C/N is a constant 30 g C/g N. For DOB 5.0 mg/l, the C/N varies from 80 g C/g N at 0 mg O2/l to 30 g C/g N at 5 mg O2/l (Parnas, 1975). If the wetland is carbon-limited, the individual amounts of POC and DOC utilized will be based on the relative fraction of each (i.e. if the TOC is 30% POC and 70% DOC, it is assumed 30% of the TOC utilized by heterotrophs will be in the particulate form and 70% of the TOC utilized will be in the dissolved form). A corresponding fraction of PON and DON will be utilized, based on the fraction of total organic carbon to total organic nitrogen in the wetland. Excess nitrogen is wasted as ammonium. If the wetland is nitrogen limited, the reverse relationships hold. Excess carbon remains as an available energy source and the wetland becomes more nitrogen-limited. The conversion of ammonium to nitrate is carried out by autotroph bacteria and the amount of ammonium utilized is a function of autotroph growth and stoichiometry. Nitrate is used by anaerobic heterotrophs as an electron acceptor, converting NO− 3 to N2 gas, which is then lost to the atmosphere. Some POC and associated PON are considered refractory and are permanently lost to the soil through peat accumulation. Peat accumulation is assumed to occur at a constant rate of 2.4 mm/year (Kadlec, 1989b; Gidley, 1995).

2.2. Oxygen budget Oxygen is added to the wetland by influent wastewater and by plants. Oxygen transfer with the atmosphere is assumed negligible compared to other oxygen sources and sinks, since the free water surface is below the substrate in subsurfaceflow wetlands. It is assumed there is a uniform vegetation stand and that plants transport oxygen to their roots at a constant rate during the growing season. This is based on the theory that processes such as Knudsen diffusion are responsible for rootzone aeration (Grosse, 1991). Oxygen in wetlands is consumed by many processes including heterotrophic and autotrophic respiration and chemical oxidation of reduced iron and manganese and sulfides (Reddy and Patrick, 1983). Microbial respiration is proportional to microbial growth; 1.23 g O2/g cells produced are used by heterotrophs and 11.9 g O2/g cells produced are used by autotrophs, based on stoichiometry. Since most constructed wetlands have gravel substrates and secondary wastewater is generally aerobic, reduced iron and manganese and sulfide concentrations are considered negligible.

2.3. Autotroph dynamics The Autotroph Dynamics submodel describes changes in the state variable NITROSOMONAS. Changes in NS population are due to growth and death of the bacteria. Growth rate is described using Monod dual substrate limitation kinetics:



m= mmax·



NH4 DO · NH4 + KNH4 DO+ KDO



(1)

where m is the actual NS specific growth rate (day − 1), mmax is the maximum growth rate (day − 1), NH+ is the ammonium concentration 4 (mg/l), KNH4 is the NS ammonium half saturation constant (mg/l), DO is the wetland dissolved oxygen concentration (mg/l), and KDO is the NS dissolved oxygen half saturation constant (mg/l). This expression modifies the NS maximum growth rate when oxygen (the electron acceptor) or ammonium (the electron donor) is limiting. The half saturation constants represent substrate

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

concentrations at which growth is half the maximum rate. The model assumes pseudo steady state conditions, based on work by Benefield and Molz (1984) to model biofilm growth on an idealized flat plate. This is based on the idea that a dynamic environment can be approximated as short, consecutive periods of steady-state conditions. In addition to substrate limitations, microbial growth is limited by temperature and pH. The optimum temperature range for NS is from 15 to 35°C (Bender, 1976; Fyock, 1977; Reddy and Patrick, 1983; Wheaton et al., 1991). Growth still occurs outside this range, though it is reduced. To approximate these effects, the NS Temperature Factor is included. This parameter is 1.0 in the optimum temperature range and then decreases linearly to 0.0 at 0°C (lower limit) and at 40°C (upper limit). The optimum pH range for NS is 6 – 9 (Fyock, 1977; Kholdebarin and Oertli, 1977; Reddy and Patrick, 1983). Because wetlands tend to drive pH towards neutrality, pH is not modeled. NS growth is the product of the actual growth rate, the NS Temperature Factor and the mass of NS bacteria in the wetland. Since there is little information about the factors affecting microbial die-off, it is modeled as a first-order reduction.

2.4. Heterotroph dynamics As with the Autotroph Dynamics submodel, Heterotroph Dynamics describes the changes in heterotrophic microbes. Nitrate is assumed to be the only other available electron acceptor when DO concentrations drop below 1 – 2 mg/l. Other electron acceptors, such as sulfate, iron, and manganese are not modeled. The Monod models are: m = mmax·





for aerobic heterotrophs, and m= mmax·





(2)



(3)

TOC DO · TOC+KTOC DO +KDO



TOC KDO · TOC+KTOC DO + KDO

523

for anaerobic heterotrophs. The maximum aerobic growth rate or maximum anaerobic growth rate is mmax (day − 1), TOC is the total organic carbon concentration (mg/l), KC is the organics half saturation constant (mg/l), DO is the dissolved oxygen concentration (mg/l), and KDO is the HT DO half saturation constant (mg/l). As DO concentrations decrease, some heterotrophs can use nitrate as an electron acceptor instead of oxygen. This representation allows a gradual switch from aerobic to anoxic respiration with neither reaching zero. Even in aerobic wetland systems, denitrification can occur in anaerobic microsites, while in anoxic or anaerobic substrates, plant roots provide oxygen for aerobes. The optimum temperature and pH ranges for HT are similar to that for NS, so the same temperature function is employed and pH effects are not modeled. HT death is modeled as a first-order loss, similar to NS death.

2.5. Water budget The water budget consists of a single state variable WATER VOLUME. Water inputs to the wetland are from wastewater inflows and precipitation. Water loss from the wetland is due to outflow and evapotranspiration (ET). The cell is assumed to be lined, so there are no groundwater exchanges. Flow through gravel beds is best modeled with Darcy’s equation (Kadlec, 1986c, 1989a; Jorgensen et al., 1988). Given the water volume in the wetland at each time step, the water depth is determined. The hydraulic gradient is then assumed to be the maximum of the bed slope or the difference in elevation between the water surface in the wetland and the outflow pipe height, since precipitation and evapotranspiration affect the entire wetland equally and some wetlands have a flat bottom. Darcy’s law assumes steady state conditions. While flow through constructed wetlands is often unsteady due to precipitation and ET, it is assumed that flow would be steady over short periods. Additionally, Darcy’s law is the only simple model for flow through porous materials (Reed et al., 1995). It is also assumed that surface flow would not occur, i.e. the gravel depth is

524

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

sufficient to maintain subsurface flow. Porosity is considered constant; the effects of peat accumulation on hydraulic conductivity and porosity are not modeled. Thornthwaite’s method is used to estimate evapotranspiration. Daily ET is approximated by using the daily average air temperature instead of the monthly average air temperature and then dividing by 30 days.

2.7. Sensiti6ity analysis A sensitivity analysis was conducted by varying each parameter by an order of magnitude lower and higher. Those parameters which are fractions were input as 0.1, 0.5, and 0.9. Model sensitivity was determined qualitatively by graphing model output from each run and noting relative changes in the output.

2.6. Mayo wetland Data from an existing constructed wetland at the Mayo Water Reclamation Facility in Anne Arundel County, MD were chosen to test the model. This experimental wastewater treatment facility was designed to treat septic tank effluent from failing drain fields. Septic tank effluent (300 – 500 m3/day) is applied to sand filters, followed by horizontal subsurface-flow emergent wetlands, vertical flow peat wetlands, and UV radiation, after which it is discharged to the Chesapeake Bay through a diffuser pipe in cultured submerged aquatic vegetation. The emergent wetland cells have dimensions of 70.1× 45.7 × 0.8 m. The available data were surveyed and 350 consecutive days of data from July 1, 1991 to June 16, 1992 were selected to test the wetland model. This time period was chosen because all of the nitrogen cycle − components (NH+ 4 -N, TKN, and NO3 -N) were measured over almost an entire year. An entire year was simulated to evaluate model performance during all four seasons. Model validation was not conducted due to the lack of additional data of similar detail. Wastewater flow, dissolved oxygen concentrations, and temperature were measured daily, while total suspended solids (TSS), BOD5, and nitrogen samples were taken biweekly by plant personnel. Where site data were not taken daily, linear interpolations were made between input data points to provide daily input data. Climate data were obtained from NOAA (National Oceanographic and Atmospheric Administration) records and Climates of the States (Ruffner, 1985) for Annapolis, MD. Parameters that were not measured at the wetland were estimated based on information from literature. For further details, see Gidley (1995).

3. Results and discussion

3.1. Model calibration To evaluate model performance, model output of outflow, BOD5, nitrogen concentrations, and dissolved oxygen were compared to actual effluent values. Parameters were adjusted to graphically fit model output to actual site data. Leaching produced a large spike in the BOD5 output that was not evident in site data. It appears the assumption that 15% of standing dead carbon leaches to DOC following the first frost is incorrect. A corresponding increase in effluent organic nitrogen concentrations did not occur. To improve the BOD5 prediction, leaching was set to zero. Several microbial parameters were lowered by one or more orders of magnitude, compared to literature values. The most dramatic changes were made to the aerobic and anaerobic heterotroph maximum growth rates, which were reduced three orders of magnitude (6.0 to 0.015 day − 1 and 4.0 to 0.014 day − 1, respectively) to reproduce site BOD5, NH+ 4 -N, and TKN effluent concentrations. Additionally, heterotroph and autotroph yields (0.6 g cells/g C and 0.2 g cells/g NH+ 4 -N) and death rates (0.05 and 0.1 day − 1) and the autotroph maximum growth rate (1.0 day − 1) were each lowered an order of magnitude. The initial mass of heterotrophs was increased slightly to improve initial dissolved oxygen and heterotroph population responses. To improve model effluent nitrate concentration predictions, the heterotroph nitrate yield was decreased from the stoichiometric value of 3.29 to 0.2 g microbes/g NO− 3 -N.

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

The literature values are typical for conventional wastewater treatment systems, so it is not unreasonable that microbial growth is slower in a less controlled environment like constructed wetlands. Lower maximum growth rates could also indicate that microbial growth is inhibited by a factor other than nutrients, such as diffusion. Assuming the design flow of 510 m3/day and a water depth of 0.75 m, the theoretical pore velocity is 50 m/day, or 0.059 cm/s. Velocities measured in conventional fixed-film nitrification systems range from 2.4 to 59.1 cm/s (Kugaprasatham et al., 1992). Similar velocities were used by Benefield and Molz (1984) to simulate organics removal by attached heterotrophs. The low velocity may not allow sufficient nutrient flux to support a large microbial substrate utilization rate. The biomass rootzone oxygenation rate was decreased an order of magnitude compared to literature values. Based on literature, the wetland biomass was assumed to translocate 5 g/m2 day to the rootzone during the growing season (Reed et al., 1995). This oxygenation rate resulted in overpredicted dissolved oxygen concentrations during the growing season. To better estimate site data, the biomass oxygenation rate was reduced to 0.4

525

g/m2 day. Rootzone oxygenation rates have been reported as low as 0.02 g/m2 day (Brix, 1997). This rate may be low because the gravel bed depth (0.8 m) is twice the rooting depth of many wetland plants. A study of constructed wetlands with similar bed depths reported reduced treatment due to short circuiting below the rootzone (Gersberg et al., 1984). A time step of 90 min was used. This produced results similar to shorter time steps while maintaining numerical stability in the model.

3.2. Model predictions In general, the results produced by the model were as expected. In the summer when wastewater temperatures are greater (24°C average in July vs. 4°C average in January), microbial populations are high and effluent concentrations are low. In the winter the opposite holds. Nitrification is reduced when oxygen concentrations and temperatures are low, while denitrification increases. The strong interdependence of the carbon, nitrogen and oxygen cycles is evident: errors in one submodel are propagated through other submodels. Results of specific submodels are discussed in more detail below.

Fig. 1. Comparison of wastewater inflow, observed outflow and predicted outflow rates. Note: entire data set not plotted due to sporadic data.

526

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

Fig. 2. Influent, observed effluent, and predicted effluent 5-day BOD concentrations.

3.2.1. Outflow The model did a fair job of simulating the volumetric outflow rate (Fig. 1). Differences in model data and actual site data could be the result of changes in the outflow pipe height by plant personnel, differences in rainfall amounts between Annapolis and the Mayo Peninsula, and errors in bed porosity estimates. 3.2.2. Biochemical oxygen demand From the graph of predicted versus actual BOD outflow (Fig. 2), it can be seen that the model generally overpredicted BOD concentrations, though the overall trends were well represented for most of the year. Model results improved when leaching was removed. The effluent BOD spike that occurred in November 1991 is not the result of increased influent BOD. Air temperatures decreased significantly in October and this spike could represent the actual death and leaching of vegetation at the facility. Because this response was not evident in the nitrogen effluent data, and no data were taken by facility personnel on vegetation growth, leaching was considered zero in the model calibration. Future studies with more complete data may reveal that leaching of dying vegetation is a significant nutrient source. Differences may also be the result of many simplifications made in the carbon cycle submodel. Influent BOD, dead microbial cells and

plant litter were lumped into POC and DOC with the assumption that these three fractions were equally available for microbial degradation. While BOD and microbial turnover are fairly constant throughout the year, plant litter inputs occur predominately in the late fall and it is well documented that plant litter becomes more resistant to degradation with time. There could be an initial leaching of labile carbon from vegetation in the fall, followed by an increasing rate of peat accumulation as the most available fractions are degraded. The sudden increase in BOD in the middle of January is the result of increasing BOD inflow concentrations on January 1, 1992. In 1992, operation of the constructed wetland was changed: 25% of the facility influent was bypassed around the initial sand filters to increase carbon in the wetland. This caused an eightfold increase in BOD load to the wetland. Because this was done in the winter when microbial growth is slow due to cold temperatures, the wetland was probably unable to assimilate this sudden load increase. Discrepancies between model results and actual BOD concentrations during 1992 could be the result of changes in model parameters due to modifications of operational procedures at the facility. Since part of the wetland influent bypassed the sand filters in 1992, the ratio of particulate to dissolved BOD probably increased.

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

Settling of particulate BOD could be the reason that actual effluent BOD concentrations were initially low after the BOD load to the wetland increased. This would explain the overestimation of model effluent BOD at this time, since model effluent BOD is assumed to be only dissolved organic carbon. The basic assumption that the wetland acts as a CSTR appears to have a significant effect on model results. When comparing model output and site data, it is evident that the model smooths

527

peaks and valleys in the effluent concentrations. This damped response could be improved by modeling the wetland as a series of smaller CSTRs instead of a single large CSTR (Kadlec, 1994). This would better approximate the nonideal plug flow observed in other constructed wetlands (Reed, 1995).

3.2.3. Nitrogen The effluent nitrogen series was well repro− duced. Predicted NH+ 4 , TKN and NO3 effluent

Fig. 3. Influent, observed effluent, and predicted effluent ammonium nitrogen concentrations.

Fig. 4. Influent, observed effluent, and predicted effluent total Kjeldahl nitrogen concentrations.

528

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

Fig. 5. Influent, observed effluent, and predicted effluent nitrite/nitrate nitrogen concentrations.

concentrations were generally close to actual effluent concentrations and seasonal trends were reproduced (Figs. 3 – 5). ON was slightly underpredicted by the model. This error may be due to the assumption that all particulates are retained in the wetland. Effluent ON probably has a small particulate fraction as the result of microbial sloughing and plant litter degradation. The most obvious errors in effluent NH+ 4 and TKN predictions are from January to May, when concentrations were overpredicted. These have a similar form as the BOD overestimation and probably are a result of increases in the wetland BOD particulate fraction and assumptions inherent to the CSTR model, as discussed above. Additionally, effluent DO was underestimated in the spring and this error could have propagated as reduced nitrification and aerobic organics degradation. Other similarities between effluent NH+ 4 and BOD concentrations are evident. In January, the model predicts a sharp increase in effluent NH+ 4 concentrations, just as with effluent BOD concentrations. Actual wetland effluent does not show a response to the increased load until the middle of January. For BOD, this is attributed to solids removal: Influent total suspended solids (TSS) concentrations increased by the same proportion as influent BOD5 in January. For ammonium, however, this delay may be the result of

ammonium sorption to the solids. Since ammonium sorption was not modeled, this effect would not have been reproduced. Model NO− predictions generally agree well 3 with site data. In the summer and fall, the model overpredicts effluent NO− concentrations, 3 whereas in late spring, concentrations are underpredicted. Denitrification appears to be more extensive in the warm months than predicted. The proportion of anaerobic heterotrophs may be higher in the summer and fall than the model predicts due to the occurrence of anaerobic microsites. Since the oxygen saturation concentration decreases with increasing water temperature, anaerobic microsites would be more prevalent during the warm summer months and the proportion of anaerobic heterotrophs would increase. Inconsistencies between model results and site data in May 1992 are probably due to errors in the DO submodel. During this period DO is significantly underestimated, probably resulting in an overestimation of denitrification. The dependence of denitrification on carbon is demonstrated in the results from December. At this time, effluent BOD concentrations are very low, which limits the carbon available to anaerobes, reduces denitrification, and causes an increase in effluent NO− 3 . This response is reflected in both actual and model effluent NO− 3 concentrations.

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

The interdependence of the carbon, nitrogen, and oxygen cycles can be observed by comparing Figs. 2–6. Overestimates of effluent NH+ 4 concentrations correspond to underestimates of effluent NO− 3 concentrations. These errors are also related to inconsistencies in effluent DO predictions. Fig. 6 shows that effluent DO was underestimated in August 1991 and throughout the spring of 1992. This corresponds to overestimates of effluent NH+ and BOD concentrations, and underesti4 mates of effluent NO− 3 concentrations. It appears that model nitrification and aerobic organics degradation were inhibited due to the low dissolved oxygen. Because there was sufficient carbon and lower DO at these times, denitrification was overestimated, resulting in an underprediction of effluent NO− 3 concentrations.

3.2.4. Dissol6ed oxygen Model results and site data are presented in Fig. 6. Because the wastewater was initially sprayed onto sand filters, wetland influent dissolved oxygen concentrations were often high. Model predictions matched site data well during the summer and fall. The large peak in predicted DO concentration in November is the result of a sudden increase in actual influent DO concentrations and a decrease in heterotroph respiration caused by colder wastewater temperatures. The

529

greatest error occurs in January–March 1992 and May 1992 when effluent DO concentrations are underpredicted. Errors may be the result of the assumption that rootzone aeration occurs only during the growing season. Site data show that, while the DO increase through the wetland is reduced, some does occur in the winter, indicating that rootzone aeration may occur in winter. The underprediction of DO concentrations in May may be the result of excessive aerobic heterotroph growth, since the predicted effluent BOD concentration is too low.

3.3. Sensiti6ity analysis In general, the model is insensitive to changes in a single parameter. The model is most sensitive to changes in parameters that affect microbial growth and substrate use directly. Aerobic and anaerobic heterotroph maximum growth rates, the heterotroph death rate, and initial heterotroph cell mass have the most impact on model output. This is to be expected, since heterotrophic bacteria are responsible not only for organics degradation, but also for ammonification and denitrification. Parameters controlling autotroph growth influence ammonium outflow significantly and dissolved oxygen somewhat, but have little effect on other model results.

Fig. 6. Observed and predicted effluent dissolved oxygen concentrations (influent concentrations not plotted for visual clarity; influent concentrations ranged from 0.5 to 10.2 mg/l).

530

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

Two parameters deserve some note. First, the biomass oxygenation rate is a direct oxygen source to the wetland and has a large effect on oxygen concentrations. In turn, this influences nitrification and ammonium concentrations. Previous studies quantified this as an average rate using an oxygen balance, so the timing and mechanics of this process are still largely unknown. Secondly, even though model input was insensitive to the required microbial C/N, it could be important for other sites. Because the Mayo wetland is greatly carbon limited, this parameter had little impact. Under different conditions, the required microbial C/N could significantly influence ammonium concentrations.

3.4. E6aluation of the Mayo wetland Calibration of the model at this site resulted in process parameter values that deviated from literature values for traditional wastewater treatment systems. This helped identify design and operational problems at the Mayo wetland. First, the cell dimensions are incorrect. The wetland gravel depth is two times larger than required. This reduces contact with the oxygenated rootzone, decreasing oxygen transfer and increasing ammonification. This additional depth also decreases pore water velocities, so that nutrient flux to the microbes may limit microbial growth. Decreasing bed depth and increasing the wetland surface area could improve wetland performance. Second, the Mayo wetland appears extremely carbon-limited. This was indicated by the relative lack of model response to changes in the parameter for required microbial C/N. The predicted wetland C/N ranges from 29 on the first day of the simulation to 4 at the beginning of November. The average C/N is 12. Nitrogen removal by denitrification will not occur unless a supplemental carbon source is used. Lastly, nitrogen and carbon loading should not be increased in the winter. During this time, microbial growth is the most inhibited and little treatment will occur. This is indicated in both actual and modeled wetland responses to an increase in influent BOD in January 1992 (Fig. 2).

4. Conclusions This model is an initial attempt to incorporate microbial dynamics in a constructed wetland simulation model. Model calibration resulted in a significant reduction in microbial growth rates and yields, indicating that processes such as diffusion may control wastewater treatment in the Mayo wetlands. Following calibration, the model reproduced seasonal trends in nitrogen, oxygen, and BOD5 concentrations well. Interactions between these cycles were evident in model output. The development of this model also identified several assumptions and parameters that require further investigation to improve constructed wetland design. These are listed below: 1 What are the effects of pore water velocities on microbial growth and treatment processes, and are minimum pore velocities necessary for design criteria in addition to hydraulic residence time? 2 Is ammonium sorption significant for delaying ammonium release and is sorbed ammonium utilized by microbes? 3 Is a simple carbon cycle adequate? Should biomass growth and nutrient uptake also be modeled with Monod kinetics? Is it appropriate to lump microbial biomass, plant litter and influent organics together, or should the decomposition of these three sources be considered separately? 4 What is the magnitude and timing of biomass rootzone aeration? Additional work should be conducted to test and improve the model before using it for design purposes.

Acknowledgements This material is based upon work supported under a National Science Foundation Graduate Fellowship. Any opinions, findings, conclusions or recommendations expressed in this publication are those of the authors and do not necessary

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

reflect the views of the National Science Foundation. Additional research funds were provided by the USDA Southern Regional Aquaculture Cen-

531

ter and Hoechst-Celanese Corporation through the Kenan Institute for Engineering, Technology, and Science at NCSU.

Appendix A. STELLA equations and initial parameters Outflows are listed in order of outflow priority A.1. Carbon cycle BIOMASS(t)=BIOMASS(t −dt) +(biomass – growth-biomass – death)*dt INIT BIOMASS=1400000 biomass – growth= Biomass – Growth – Rate*Surface – Area*Biomass – C – Content biomass – death= Winter*BIOMASS DOC(t)= DOC(t−dt) + (soluble – BOD – inflow+ DOC – leaching− DOC – outflow-DOC – min¯imob)*dt INIT DOC=754.0 soluble – BOD – inflow =1.4*BOD – Influent – Concentration*(1− BOD – Particulate – Fraction)*BOD – C – Fraction*inflow DOC – leaching=Surface – Area*Leaching – Rate DOC – outflow=DOC/WATER – VOLUME*outflow DOC – min¯imob = IF((TOC/TON) \ Microbe Total C:N)THEN(DON/TON*HT – growth/HT – Yield)ELSE(DOC/TOC*HT – growth/HT – Yield) POC(t)=POC(t −dt) + (particulate – BOD – inflow+ physical – degradation+ microbial – death-peat – accumulation-POC – min¯imob)*dt INIT POC=340000 particulate – BOD – inflow =1.4*BOD – Influent – Concentration*BOD – Particulate – Fraction*BOD – C – Fraction*inflow physical – degradation = Biomass – Degradation – Rate*STANDING – DEAD microbial – death =(NS – death + HT – death)*Microbial – C – Content peat – accumulation=Peat – Accumulation – Rate*Peat – C – Content POC – min¯imob =IF((TOC/TON) \ Microbe Total C:N)THEN(PON/TON*HT – growth/HT – Yield)ELSE(POC/TOC*HT – growth/HT – Yield) REFRACTORY – C(t) = REFRACTORY – C(t − dt)+ (peat – accumulation)*dt INIT REFRACTORY – C =3000000 peat – accumulation= Peat – Accumulation – Rate*Peat – C – Content STANDING – DEAD(t) = STANDING – DEAD(t− dt)+ (biomass – death-physical – degradation-DOC – leaching)*dt INIT STANDING – DEAD = 860000 biomass – death =Winter*BIOMASS physical – degradation = Biomass – Degradation – Rate*STANDING – DEAD DOC – leaching= Surface – Area*Leaching – Rate Biomass – C – Content =0.47 Biomass – Degradation – Rate =0.0063 BOD – C – Fraction = 0.82 BOD – Particulate – Fraction = 0.5 Leaching – Rate = 0.0 Microbial – C – Content =0.53

532

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

Peat – Accumulation – Rate =2300 Peat – C – Content= 0.80 TOC= DOC+POC A.2. Autotroph dynamics NITROSOMONAS(t) = NITROSOMONAS(t − dt)+ (NS – growth− NS – death)*dt INIT NITROSOMONAS=1560 NS – growth = MIN((0.084*DISSOLVED – OXYGEN),(NS – Max – Growth – Rate*((NH4/WATER – VOLUME)/(NS – NH4 – Half – Sat – Constant+ (NH4/WATER – VOLUME)))*((DISSOLVED – OXYGEN/WATER – VOLUME)/(NS – DO – Half – Saturation – Constant +(DISSOLVED – OXYGEN/WATER – VOLUME)))*NITROSOMONAS*NS – Temperature – Factor)) NS – death= NS – Death – Rate*NITROSOMONAS NS – Death – Rate= 0.002 NS – DO – Half – Saturation – Constant =1.0 NS – Max – Growth – Rate =0.01 NS – NH4 – Half – Sat – Constant =1.0 A.3. Heterotroph dynamics HETEROTROPHS(t) =HETEROTROPHS(t − dt)+ (HT – growth− HT – death)*dt INIT HETEROTROPHS=250000 HT – growth= Anaerobic – HT – Growth + Aerobic – HT – Growth HT – death= HT – Death – Rate*HETEROTROPHS Aerobic – HT – Growth =MIN((1.23*DISSOLVED – OXYGEN),(Aerobic – Max – Growth – Rate*(TOC/WATER – VOLUME)/((TOC/WATER – VOLUME)+ HT – Organics – Half – Sat – Constant)*HT – Temperature – Factor*(DISSOLVED – OXYGEN/WATER – VOLUME)/((DISSOLVED – OXYGEN/WATER – VOLUME) + HT – DO – Half – Sat – Constant)*Anaerobe – Fraction*HETEROTROPHS)) Aerobic – Max – Growth – Rate =0.015 Anaerobic – HT – Growth =MIN((3.29*NO3),(Anaerobic – Max – Growth – Rate*(TOC/WATER – VOLUME)/((TOC/WATER – VOLUME) + HT – Organics – Half – Sat – Constant)*HT – Temperature – Factor*HT – DO – Half – Sat – Constant/(HT – DO – Half – Sat – Constant+ (DISSOLVED – OXYGEN/WATER – VOLUME))*(NO3/WATER – VOLUME)/((NO3/WATER – VOLUME)+ HT – NO3 – Half – Sat – Constant)*Anaerobe – Fraction*HETEROTROPHS)) Anaerobic – Max – Growth – Rate = 0.014 HT – Death – Rate =0.005 HT – DO – Half – Sat – Constant =1. HT – NO3 – Half – Sat – Constant = 0.15 HT – Organics – Half – Sat – Constant = 50 A.4. Nitrogen cycle DON(t)=DON(t −dt) +(DON – influent + N – leaching-DON – outflow-DON – immobilization-DON – mineralization)*dt INIT DON=754

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

533

DON – influent=Influent – Organic – N – Conc*(1 − ON – Particulate – Fraction)*inflow N – leaching= DOC – leaching/Biomass – C:N DON – outflow = DON/WATER – VOLUME*outflow DON – immobilization= IF((TOC/TON) \ Microbe Total C:N) THEN(DON/TON – & – NH4*Microbial – N – Content*HT – growth)ELSE(DOC/TOC*(Microbial – N – Content*HT – growth − HT – NH4 – Immobilization)) DON – ammonification = IF(((TOC/TON) \ Microbe Total C:N) OR(DOCB0.1))THEN(0.0)ELSE(DOC/TOC*HT – growth/HT – Yield*DON/DOC-DON – immobilization) IMMOBILIZED – N(t) = IMMOBILIZED – N(t − dt)+ (nitrate – uptake+ DON – immobilization + ammonium – uptake +PON – immobilization-death)*dt INIT IMMOBILIZED – N =116000 nitrate – uptake =MIN((NO3),(biomass – growth/Biomass – C:N)) DON – immobilization =IF((TOC/TON) \ Microbe Total C:N) THEN(DON/TON – & – NH4*Microbial – N – Content*HT – growth)ELSE(DOC/TOC*(Microbial – N – Content−HT – NH4 – Immobilization)) ammonium – uptake =NS – growth*Microbial – N – Content+ (biomass – growth/Biomass – C:N-nitrate – uptake)+HT – NH4 – Immobilization PON – immobilization =IF((TOC/TON) \ Microbe Total C:N) THEN(PON/TON – & – NH4*Microbial – N – Content*HT – growth)ELSE(POC/TOC*(Microbial – N – Content*HT – growth −HT – NH4 – Immobilization)) death= physical – degradation/Biomass – C:N+ Microbial – N – Content*(HT – death+ NS – death) NH4(t)= NH4(t−dt) + (NH4 – influent + DON – mineralization+ PON – mineralization-NH4 – Outflow -ammonium – uptake-nitrification)*dt INIT NH4= 0.01 NH4 – influent =NH4 – Influent – Conc*inflow DON – mineralization=IF(((TOC/TON) \ Microbe Total C:N) OR(DOCB0.1))THEN(0.0)ELSE(DOC/TOC*HT – growth/HT – Yield*DON/DOC-DON – immobilization) PON – mineralization=IF(((TOC/TON) \ Microbe Total C:N) OR(POCB 0.1))THEN(0.)ELSE(POC/TOC*HT – growth/HT – Yield)*PON/POC-(POC /TOC*(Microbial – N – Content*HT – growth − HT – NH4 – Immobilization)) NH4 – outflow=NH4/WATER – VOLUME*outflow ammonium – uptake =NS – growth*Microbial – N – Content+ (biomass – growth/Biomass – C:N-nitrate – uptake)+ HT – NH4 – Immobilization nitrification=NS – growth/NS – Yield NO3(t)=NO3(t−dt) +(NO3 – influent + nitrification-nitrate – uptake− NO3 – outflow -denitrification)*dt INIT NO3= 10500.0 NO3 – influent=NO3 – Influent – Conc*inflow nitrification=NS – growth/NS – Yield nitrate – uptake =MIN((NO3),(biomass – growth/Biomass – C:N)) NO3 – outflow=NO3/WATER – VOLUME*outflow denitrification= Anaerobic – HT – Growth/HT – NO3 – Yield PON(t)= PON(t −dt) + (death + PON – influent-peat – N – accumulation-PON – immobilization-PON – mineralization)*dt INIT PON=10600

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

534

death=physical – degradation/Biomass – C:N +Microbial – N – Content*(HT – death+ NS – death) PON – influent=Influent – Organic – N – Conc*ON – Particulate – Fraction*inflow peat – N – accumulation =Peat – Accumulation – Rate*Peat – N – Content PON – immobilization =IF((TOC/TON) \ Microbe Total C:N) THEN(PON/TON – & – NH4*Microbial – N – Content*HT – growth)ELSE(POC/TOC*(Microbial – N – Content*HT – growth-HT – NH4 – Immobilization)) PON – ammonification = IF(((TOC/TON) \ Microbe Total C:N) OR(POCB 0.1))THEN(0.)ELSE(POC/TOC*HT – growth/HT – Yield)*PON/POC − (POC/TOC*(Microbial – N – Content*HT – growth− HT – NH4 – Immobilization)) REFRACTORY – N(t) = REFRACTORY – N(t − dt)+ (peat – N – accumulation)*dt INIT REFRACTORY – N =95000 peat – N – accumulation =Peat – Accumulation – Rate*Peat – N – Content Biomass – C:N=23.5 HT – NH4 – Immobilization = Microbial – N – Content*HT – growth*NH4/TON – & – NH4 HT – NO3 – Yield= 0.2

Microbial – N – Content =0.124 NS – Yield= 0.005 ON – Particulate – Fraction =0.5 Peat – N – Content =0.025 TON= DON +PON TON – & – NH4=TON + NH4 A.5. Oxygen budget DISSOLVED – OXYGEN(t) =DISSOLVED – OXYGEN(t− dt)+ (influent – DO+ biomass – flux-DO – Outflow-HT – respiration-NS – respiration)*dt INIT DISSOLVED – OXYGEN =2300 influent – DO =Influent – DO – Conc*inflow biomass – flux =Biomass – Oxygenation – Rate*Surface – Area DO – outflow=DISSOLVED – OXYGEN/WATER – VOLUME*outflow HT – respiration =Aerobic – HT – Growth/HT – DO – Yield NS – respiration= NS – growth/NS – DO – Yield HT – DO – Yield= 1.23 NS – DO – Yield =0.084 A.6. Water budget WATER – VOLUME(t) =WATER – VOLUME(t − dt)+ (inflow+precipitation-evapotranspiration -outflow)*dt INIT WATER – VOLUME =730 precipitation= (Daily – Precipitation*2.54/100)*(Section – Length*Bed – Width) evapotranspiration =IF(Air – Temperature \ 0)THEN(1.6*Daylength*(10*Air – Temperature/Heat – Index). a)/(30*100)*(Section – Length*Bed – Width)ELSE(0) outflow= MAX(Bed – Width*Outflow – Pipe – Height*Hydraulic – Conductivity*Bed – Slope,(WATER – VOLUME*Hydraulic – Conductivity/(Section – Length. 2*Porosity)*(WATER – VOLUME/(Section – Length*Bed – Width*Porosity) -Outflow – Pipe – Height)))

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

535

a =1.476 Bed – Slope = 0.001 Bed – Width=45.7 Heat – Index=62.5 Hydraulic – Conductivity = 1000 Outflow – Pipe – Height =0.5 Porosity= 0.3 Section – Length= 70.1 Surface – Area = Section – Length*Bed – Width References Bavor, H.J., Roser, D.J., Fisher, P.J., Smalls, I.C., 1991. Performance of solid-matrix wetland systems, viewed as fixed-film bioreactors. In: Hammer, D.A. (Ed.), Constructed Wetlands for Wastewater Treatment. Lewis, Chelsea, MI, pp. 646–656. Bender, M.F., 1976. Nitrogen Dynamics and Modeling in a Freshwater Wetland. Ph.D. Dissertation, University of Michigan, Microfiche title c77-7868 390542 Film 16044 Diss. Benefield, L., Molz, F., 1984. Mathematical simulation of a biofilm process. Biotechnol. Bioeng. 27, 921–931. Brix, H., 1994. Use of constructed wetlands in water pollution control: Historical development, present status, and future perspectives. Water Sci. Technol. 30 (8), 209–223. Brix, H., 1997. Do macrophytes play a role in constructed wetland treatment? Water Sci. Technol. 35 (5), 11–17. Brown, M.T., 1988. A simulation model of hydrology and nutrient dynamics in wetlands. Comput. Environ. Urban Syst. 12, 221 – 237. Buchberger, S.G., Shaw, G.B., 1995. An approach toward rational design of constructed wetlands for wastewater treatment. Ecol. Eng. 4 (4), 249–275. Chescheir, G.M., Gilliam, J.W., Skaggs, R.W., Broadhead, R.G., 1987. The hydrology and pollutant removal effectiveness of wetland buffer areas receiving pumped agricultural drainage water, Report No. 231. Water Resources Research Institute of the University of North Carolina, August 1987. Conley, L.M., Dick, R.I., Lion, L.W., 1991. An assessment of the root zone method of wastewater treatment. Res. J. Water Pollut. Contr. Fed. 63 (3), 239–247. Costanza, R., Sklar, R.H., 1985. Articulation, accuracy and effectiveness of mathematical models: A review of freshwater wetland applications. Ecol. Model. 27, 45–68. EPA, 1993. Subsurface Flow Constructed Wetlands for Wastewater Treatment: A Technology Assessment, EPA 832-R-93-008. EPA Office of Research and Development, Washington, DC. Fyock, O.L., 1977. Nitrification Requirements of Water Reuse Systems for Rainbow Trout, No. 41. Colorado Division of Wildlife, Fisheries Research Section, February, 1977.

Gersberg, R.M., Elkins, B.V., Goldman, C.R., 1984. Use of artificial wetlands to remove nitrogen from wastewater. J. Water Pollut. Contr. Fed. 56 (2), 152 – 156. Gidley, T.M., 1995. Development of a Constructed Subsurface Flow Wetland Simulation Model. M.S. Thesis, North Carolina State University, Raleigh, NC. Grosse, W., 1991. Thermoosmotic air transport in aquatic plants affecting growth activities and oxygen diffusion to wetlands soils. In: Hammer, D.A. (Ed.), Constructed Wetlands for Wastewater Treatment. Lewis, Chelsea, MI, pp. 469 – 476. Hammer, D.E., 1984. An Engineering Model of Wetland/ Wastewater Interactions. Dissertation, University of Michigan. Hammer, D.A., 1989. Constructed Wetlands for Wastewater Treatment. Lewis, Chelsea, MI. Heliotis, F.D., DeWitt, C.B., 1983. A conceptual model of nutrient cycling in wetlands used for wastewater treatment: A literature analysis. Wetlands 3, 134 – 152. High Performance Systems, Inc, 1992. STELLA II. Hanover, NH. Howard-Williams, C., 1985. Cycling and retention of nitrogen and phosphorus in wetlands: a theoretical and applied perspective. Freshwater Biol. 15, 391 – 431. Jansson, S.L., Persson, J., 1982. Mineralization and immobilization of soil nitrogen. In: Stevenson, F.J. (Ed.), Nitrogen in Agricultural Soils. American Society of Agronomy, Inc., Crop Science Society of America, Inc., and Soil Science Society of America, Inc., Madison, WI, pp. 229 – 252. Johnston, C.A., 1991. Sediment and nutrient retention by freshwater wetlands: Effects on surface water quality. Crit. Rev. Environ. Contr. 21 (5-6), 491 – 565. Jorgensen, S.E., Hoffmann, C.C., Mitsch, W.J., 1988. Modelling nutrient retention by a reedswamp and wet meadow in Denmark. In: Mitsch, W.J., Straskraba, M., Jorgensen, S.E. (Eds.), Wetland Modelling. Elsevier, New York, pp. 133 – 151. Kadlec, J.A., 1986a. Nutrient dynamics in wetlands. In: Reddy, K.R., Smith, W.H. (Eds.), Aquatic Plants for Water Treatment and Resource Recovery. Magnolia Publishing, Orlando, FL, pp. 393 – 419. Kadlec, R.H., 1986b. Northern natural wetland water treatment systems. In: Reddy, K.R., Smith, W.H. (Eds.), Aquatic Plants for Water Treatment and Resource Recovery. Magnolia Publishing, Orlando, FL, pp. 83 – 98.

536

T.M. Wynn, S.K. Liehr / Ecological Engineering 16 (2001) 519–536

Kadlec, R.H., 1986c. The hydrodynamics of wetland water treatment systems. In: Reddy, K.R., Smith, W.H. (Eds.), Aquatic Plants for Water Treatment and Resource Recovery. Magnolia Publishing, Orlando, FL, pp. 373–392. Kadlec, R.H., 1989a. Hydrologic factors in wetland water treatment. In: Hammer, D.A. (Ed.), Constructed Wetlands for Wastewater Treatment. Lewis, Chelsea, MI, pp. 21–40. Kadlec, R.H., 1989b. Decomposition in wastewater wetlands. In: Hammer, D.A. (Ed.), Constructed Wetlands for Wastewater Treatment. Lewis, Chelsea, MI, pp. 459–468. Kadlec, R.H., 1994. Detention and mixing in free water wetlands. Ecol. Eng. 3, 345–380. Kadlec, R.H., Hammer, D.E., 1988. Modeling nutrient behavior in wetlands. Ecol. Model. 40, 37–66. Kadlec, R.H., Knight, R.L., 1996. Treatment Wetlands. Lewis, Boca Raton, FL. Kholdebarin, B., Oertli, J.J., 1977. Effect of pH and ammonia on the rate of nitrification of surface water. Res. J. Water Pollut. Contr. Fed. 49, 1689–1692. King, A.C., Mitchell, C.A., Howes, T., 1997. Hydraulic tracer studies in a pilot scale subsurface flow constructed wetland. Water Sci. Technol. 35 (5), 189–196. Knight, R.L., 1994. North American Wetland Treatment System Database. U.S. EPA, Environmental Compliance Laboratory, Cincinnati, OH. Kulshreshtha, M., Gopal, B., 1982. Decomposition of freshwater wetland vegetation. II. Aboveground organs of emergent macrophytes. In: Gopal, B., Turner, R.E., Wetzel, R.G., Whigham, D.F. (Eds.), Wetlands: Ecology and Management. UNESCO, National Institute of Ecology and International Scientific Publications, pp. 279–292. Kugaprasatham, S., Nagaoka, H., Ohgaki, S., 1992. Effect of turbulence on nitrifying biofilms at non-limiting substrate conditions. Water Res. 26 (12), 1629–1638. McCarty, P.L., 1975. Stoichiometry of biological reactions. Prog. Water Technol. 7, 157–172. Mitsch, W.J., Jorgensen, S.E., 1989. Ecological Engineering. Wiley, New York. Parnas, H., 1975. Model for decomposition of organic material by microorganisms. Soil Biol. Biochem. 7, 161–169. Patrick Jr, W.H., 1982. Nitrogen transformations in submerged soils. In: Stevenson, F.J. (Ed). Nitrogen in Agricul-

.

tural Soils. American Society of Agronomy, Inc., Crop Science Society of America, Inc., and Soil Science Society of America, Inc., Madison, WI, pp. 449-466. Polunin, N.V.C., 1982. Processes in the decay of reed (Phragmites australis) litter in fresh water. In: Gopal, B., Turner, R.E., Wetzel, R.G., Whigham, D.F. (Eds.), Wetlands: Ecology and Management. UNESCO, National Institute of Ecology and International Scientific Publications, pp. 293 – 304. Reddy, K.R., D’Angelo, E.M., 1997. Biogeochemical indicators to evaluate pollutant removal efficiency in constructed wetlands. Water Sci. Technol. 35 (5), 1 – 10. Reddy, K.R., Patrick, W.H., 1983. Nitrogen transformations and loss in flooded soils and sediments. CRC Crit. Rev. Environ. Cont. 13 (4), 273 – 309. Reed, S.C., 1990. Natural systems for wastewater treatment. In: Water Pollution Control Federation Manual of Practice FD-16. Reed, S.C., 1995. Design of subsurface flow constructed wetlands for wastewater treatment. In: Reed, S.C., Middlebrooks, E.J., Crites, R.W. (Eds.), Natural Systems for Waste Management and Treatment, 2nd edn. McGrawHill, New York. Reed, S.C., Brown, D., 1992. Subsurface flow wetlands — A performance evaluation. In: Water Environment Federation 65th Annual Conference and Exposition, New Orleans, LA, September 20 – 24, Vol. IX, pp. 3 – 11. Reed, S.C., Middlebrooks, E.J., Crites, R.W., 1995. Natural Systems for Waste Management and Treatment, 2nd edn. McGraw-Hill, New York. Ruffner, J.A., 1985. Climates of the States, vol. 1. Gale Research, Detroit, MI. Watson, J.T., Reed, S.C., Kadlec, R.H., Knight, R.L., Whitehouse, A.E., 1991. Performance expectations and loading rates for constructed wetlands. In: Hammer, D.A. (Ed.), Constructed Wetlands for Wastewater Treatment. Lewis, Chelsea, MI, pp. 319 – 352. Wheaton, F., Hochheimer, J., Kaiser, G.E., 1991. Fixed film nitrification filters for aquaculture. In: Brune, D.E., Tomasso, J.R. (Eds.), Aquaculture and Water Quality. The World Aquaculture Society, Baton Rouge, LA, pp. 272 – 303.