adaptation of wastewater surface flow wetland formulae for application in constructed stormwater...

Ecological Engineering 9 (1997) 187–202

Adaptation of wastewater surface flow wetlandformulae for application in constructed

stormwater wetlands

T.H.F Wong a,*, W.F. Geiger b

a Department of Ci6il Engineering, Monash Uni6ersity, PO Box 197, Caulfield East,Victoria 3145, Australia

b Department of Ci6il Engineering, Uni6ersity of Essen, Uni6ersitatstrasse 15, Essen 45141, Germany

Received 14 February 1997; received in revised form 22 August 1997; accepted 25 August 1997

Abstract

Over the past 30 years, the use of constructed wetlands for wastewater treatment has beena topic of significant research culminating in a good data base from which simplisticequations have been derived to aid in the design of these facilities to meet long term waterquality treatment performance criteria. Over the past decade, the use of treatment wetlandshas extended to stormwater and combined sewer overflow (CSO) management applications.Designing constructed wetlands for stormwater and CSO applications have unique chal-lenges stemming from the highly stochastic nature of the hydraulic and pollutant loading ona stormwater wetland compared with wastewater treatment systems. This paper explores thepossibility of adapting the simplistic models for wastewater wetlands for interim use indeveloping design guidelines for stormwater wetland systems. A procedure that takes intoaccount the unsteady intermittent nature of stormwater inflows to these wetlands has beenincorporated into one of these simplistic models and a case study presented to demonstratethe application of the procedure. © 1997 Elsevier Science B.V.

Keywords: Constructed wetlands; Wastewater; Stormwater; Stochastic processes; Hydrologiceffectiveness

* Corresponding author. Tel.: +61 39 9032557; fax: +61 39 9031019;e-mail: [email protected]

0925-8574/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved.

PII S0925 -8574 (97 )10011 -8

T.H.F. Wong, W.F. Geiger / Ecological Engineering 9 (1997) 187–202188

1. Introduction

The use of constructed wetlands for wastewater treatment has proven to besuccessful since its first serious introduction as water treatment technology in theearly 1970s. Over the past 20 years, much data has been collected on the effective-ness of wetlands in the treatment of a number of key water quality parameterincluding TSS (total suspended solid), BOD (biochemical oxygen demand), TP(total phosphorus) and TN (total nitrogen). Analysis of this data by manyresearchers has found the suitability of first order decay functions in representingthe rate of concentration reduction in the water quality parameter. A typical modelis a two-parameter first order decay function, which expresses the rate at whichpollutant concentration decreases with distance along the wetland as a linearfunction of the concentration. The model assumes steady and plug flow conditionsand is typically expressed as follows:

qdCdx

= −k (C−C*) (1)

where

q hydraulic loading rate (m/yr), defined as the ratio of the inflow and thesurface area of the systemfraction of distance from inlet to outletx

C concentration of the water quality parameterbackground concentration of the water quality parameterC*

k areal rate constant (m/yr)

Over the past 20 years, monitoring of constructed wastewater wetlands have, inpart, been devoted towards determination of the appropriate values of the rateconstants (k) and background concentrations (C*) for a number of key waterquality parameters.

Since the late 1980s, the concept of wetland treatment has extended to treatingstormwater runoff and combined sewer overflows (CSO) from urban areas. Theconditions at which these stormwater and CSO systems function are significantlydifferent from wastewater treatment systems, most notably in the stochastic natureof stormwater and pollutant inputs to these systems. Research into the performanceof stormwater wetlands is still in its infancy and there are currently limitedguidelines available on the proper sizing of these facilities to meet water qualityobjectives (Somes and Wong, 1994).

It is possible to adapt current simplistic models on water quality performance ofwastewater wetlands for use as an interim procedure for designing stormwaterwetland systems. To do this, it is first necessary to identify those conditions thatdifferentiate their characteristics.

T.H.F. Wong, W.F. Geiger / Ecological Engineering 9 (1997) 187–202 189

2. The ‘k-C* Model’

2.1. Background

The model used in predicting the performance of wastewater wetlands asgenerally expressed in Eq. (1) is referred to by Kadlec and Knight (1996) as the‘k-C* model’. The model involves two parameters, i.e. the rate constant k and thebackground concentration C*.

Integrating Eq. (1) over the length of a wetland gives the equation

Co−C*Ci−C*

=e−k/q (2)

where Ci and CO are the concentrations at the inlet and outlet of the wetlandrespectively. The treatment mechanism modelled in Eq. (2) is based on the conceptof surface contact. The ratio of the areal rate constant to the hydraulic loading(k/q) determines the effectiveness of the particular system. An alternative approach,based on the concept of water detention can also be formulated in which theproduct of the volumetric rate constant and the nominal detention period replaces(k/q) in the above equation (Kadlec, 1996). The areal rate approach appears to bemore relevant to the observed behaviour of wastewater wetlands as evident by thesignificantly larger database of k values for wetlands.

2.2. Design equations

The expression for computing the area of wetland required to meet a givenoutflow concentration criterion can be obtained by relating the inflow rate (Q) tothe unit area hydraulic loading q as follows:

Q [m3/y]=A [m2]×q [m/y] (3)

combining with Eq. (2) gives:

Co−C*Ci−C*

=e−kA/Q (4)

and

A=Qk

ln�Ci−C*

Co−C*�

(5a)

or

Co=C*+ (Ci−C*) e−kA/Q (5b)

Eq. (5a) and Eq. (5b) can be used to either determine the appropriate size of aconstructed wetland or to determine the expected effluent concentration of anexisting wetland for different influent pollutant concentrations.


2.3. Design parameters for wastewater wetlands

Typical values of the areal rate constant and background concentrations fordifferent water quality parameters derived by researchers have been collated andpresented by Kadlec and Knight (1996) and reproduced in Table 1. There is a lotof uncertainty in the values of the two parameters k and C* listed in Table 1. Thisis a reflection of a combination of intersystem variability attributed to differences inwetland physical and ecological characteristics as well as catchment and pollutantcharacteristics. Furthermore, the assumptions of plug flow and constant parametervalue are also not strictly correct. The influence of such factors as wetland depth,shape of the wetland, inlet and outlet locations, vegetation type and density, soiltype, level of mixing within the wetland etc. can be expected to contributesignificantly to the variability of the parameters evident in Table 1.

In spite of the large variation in the parameter values of k and C*, empirical dataof individual wetlands on the spatial distribution of pollutant concentrations havetended to confirm the applicability of the general form of Eq. (1). This wouldsuggest that the inadequacies may not be in the form of the model but rather in theproper definition of its parameters to reflect catchment and wetland characteristics.It can be expected that the uncertainty of the parameters would be significantlyreduced when local data becomes available for calibration of the model eitherthrough a pilot scheme or performance monitoring during the early stages of the

Table 1Rate constants and background concentrations for key water quality parameters

k (m/yr); C*Water quality Remarksparameter

Represents the settling velocity of the particles in thek=1000 m/yrTotal suspendedsolids water column. Empirical data gives the range of k to

be between 1000 and 10 000 m/yr.The relationship for the background concentrationC*=5.1+0.16Ci

of TSS is poorly defined with an R2 value of 0.23with 1582 data points.

Biochemical oxygen Average of data from 20 wetlands at eight systemsk=34 m/yrdemand with a standard deviation of 22. Range is between

6.5 and 93.7.C*=3.5+0.053Ci The relationship for the background concentration

of BOD has an R2 of 0.67 based on 33 data points.(avg.=6.2 for marsh; 1.9 for forested wetlands)

Total phosphorus Based on 20 emergent marsh systems with a stan-k=12 m/yrdard deviation of 6.1.

C*=0.02 mg/l General lower limit

k=22 m/yrTotal nitrogen Based on 59 wetland systems with rates rangingfrom 0.56 to 66.

C*=1.50 mg/l General lower limit


operation of the constructed wetland. Regional design guidelines, which progres-sively incorporate local experiences on the performances of wetlands, can bedeveloped on this basis.

The k values listed in Table 1 are areal rate constants and inherently assumeindependence of depth although the operating depth range of most typical wetlandsis not expected to be large. Kadlec and Knight (1996) have adopted a notionalmean depth of 0.3 m with a range of between 0.15 m and 0.45 m in theircalculations of the hydraulic retention time.

Stormwater wetlands are subject to a wide range of hydraulic loading and servemultiple functions. As a consequence, their depth range has tended to be wider thanwastewater wetlands. Stormwater wetlands are also different in their operationfrom stormwater quality control ponds in that the depth range needs to besufficiently wide to support a variety of wetland vegetation types. Somes et al.(1996), in their discussion on integrating hydrological and botanical design consid-erations, examines the suitable depth of inundation and frequency of wetting anddrying necessary to support a diverse vegetation characteristics within the wetland.Typical depth range examined in that study was up to 1 m above the permanentpool level. Typically the depth of the permanent pool is approximately 0.3 m andsimulations by Somes et al. (1996) for Melbourne conditions found water depth tobe between 0 and 0.6 m above the permanent pool for 80% of the time owing to thestochasticity of stormwater inflow.

2.4. Incorporating stochastic inputs

The application of the ‘k-C* model’ assumes steady inflow both in terms of flowquantity and pollutant concentration. Departures from steady flow conditionscould lead to differences between predicted and observed pollutant concentrations.The model is most suited for prediction of long-term water quality performance ofwastewater wetlands. A number of elements of stochasticity in the wastewatertreatment train can lead to incorrect prediction of pollutant concentrations andloads over short time intervals (weeks instead of months). Kadlec (1996) suggestedthe main element of stochasticity to be:� influence of precipitation, evapotranspiration and infiltration on the background

concentration and the hydraulic loading of the wetland.� influence of temperature variation and seasonality on the biological processes in

the wetland.In stormwater wetlands, the level of stochasticity increases owing to the followingcharacteristics of stormwater runoff.1. Unsteady intermittent nature of runoff resulting from variability of rainfall

depth, storm duration and storm temporal pattern.2. Unsteady intermittent nature of pollutant loading resulting from stochasticity of

runoff discussed above and variability in the rate of pollutant accumulationduring the period preceding a storm event and variable time-distribution ofpollutant concentration during a storm event.


3. Different time-distribution characteristics of pollutant concentration dependenton the water quality parameter in question during a storm event.

In attempting to adapt the ‘k-C* model’ for stormwater application, it will benecessary for the possible effects of the above listed stochasticity to be taken intoaccount. The following sections examines the characteristics of these effects andpossible means of incorporating them into the ‘k-C*’ model.

3. Unsteady intermittent hydraulic and pollutant loading

The intermittent nature of rainfall and the variability in the rainfall depth, stormduration and temporal pattern will produce unsteady intermittent hydraulic loadingto stormwater wetlands. The unsteady intermittent pollutant loading of stormwaterwetlands further complicates the stochasticity of these systems, as pollutant concen-trations are not necessarily correlated with discharge. Large events, while carryinghigh pollutant loads, have often not been found to have the highest pollutantconcentration (Geiger, 1984).

3.1. Unsteady intermittent inflows

With near steady inflow, the size of a constructed wastewater wetland may beselected by specifying the desired detention period with the expectation that a highpercentage of the inflow would be subjected to this desired period of detentionwithin the wetland. The effectiveness of stormwater treatment by detention(whether by wetlands or detention basins) is conditional on the antecedent waterlevel in the detention system. The antecedent water level immediately prior to theoccurrence of stormwater or CSO inflows to the detention system is dependent onthe available detention storage volume, the emptying rate of the detention systemand the period between storm events. Geiger (1987) adopted a continuous simula-tion approach to evaluate the significance of these factors in the design of detentionbasins for CSO applications. In a study using a similar approach, Wong and Somes(1995) derived interaction charts highlighting the relationship between three keyparameters defining the operation of a wetland with an extending detention basinunder intermittent flow conditions, i.e.,1. The detention time of the wetland.2. The volume of extended wetland storage (i.e. above the permanent pool level)

available for inflow detention.3. The overall percentage of runoff (hydrologic effectiveness) which can be ex-

pected to be subjected to the treatment processes in the wetland.The above three key parameters are interrelated in that for a given size constructedwetland, the hydrologic effectiveness varies inversely with the detention time of thebasin. Fig. 1 shows the hydrologic effectiveness curves developed for constructedwetlands in Melbourne, Australia. The curves were derived from continuoussimulations of the behaviour of storages in the Melbourne region using generatedrunoff from 100 years of continuous rainfall data recorded in Melbourne. The


Fig. 1. Hydrologic effectiveness of wetlands in Melbourne (Wong and Somes, 1995).

simulations examine the amount of runoff being diverted away, and bypassing, thebasin owing to it being full. In all cases, the discharge control of the basin is by ariser structure with flow proposition being first in/first out.

The hydrologic effectiveness curves shown in Fig. 1 are unique to the Melbourneregion and will be different from one region to another owing to differences in thecharacteristics in their respective rainfall intensity–frequency–duration relation-ships, seasonal rainfall distribution, rainfall durations and inter-event dry periods.

3.2. Unsteady intermittent pollutant loading

A combination of many factors influences the temporal characteristics ofstormwater and CSO pollutants (i.e. the pollutograph) generated from stormevents. Some of these factors are stochastic in nature often related to the meteoro-logical characteristics such as the temporal and areal distribution of rainfall overthe catchment, the duration of dry periods between storm events, etc and withhuman activities within the catchment, e.g. litter, pet droppings, car washing etc.Other factors are more systematic and are related to catchment shape, landuse,areal distribution of pollutant sources etc. It is therefore not surprising to note thatthere is often no clear trend in the shape of the pollutographs from one catchmentto another. Within a given catchment, there may be some general tendencies for theshapes of the pollutographs to have common characteristics in timing and magni-tudes of pollutant concentrations. The strength of this trend is largely dependent onthe relative dominance of systematic factors over stochastic factors in pollutantgeneration.

In an investigation of pollutant loading in a combined sewer system in Munich,Geiger (1984) computed curves representing the relationship between cumulatedpollutant load with cumulated runoff volume for a range of pollutants in an


attempt to classify the characteristics of their pollutographs. The shapes of thecumulative curves were categorised into the three broad groups exhibiting earlyflush, uniform and delay washoff pollutant transport characteristics. Analysis of theshape of observed pollutographs were carried out for six water quality parametersfor the combined sewer system in Munich and the results are summarised in Table2.

The catchment analysed by Geiger (1984) is typically flat and the pollutantgeneration mechanism is expected to be significantly influenced by systematicfactors related to catchment shape, landuse distribution and sewer network charac-teristics and the stochastic influence of meteorological factors is expected to beattenuated. These systematic influences were considered to be conducive to theoccurrence of first flush in pollutant transport with the tendency for the polluto-graph to peak before the hydrograph. This was largely confirmed by the data,which showed that over 60% of the pollutographs analysed exhibited a tendency ofan early flush mechanism. However, the remaining events showed a direct reversalof this trend with about 30% of the events showing a tendency for pollutanttransport to be delayed compared with corresponding runoff hydrographs.

Without detailed and substantial amounts of monitoring and analysis of waterquality and quantity data, it would not be possible to ascertain the relativesignificance of the stochastic and systematic factors which can affect the pollutantgeneration characteristics of a catchment. Even with available data, as was the casewith the analysis of Geiger (1984), it is unlikely that the trend can be completelysubstantiated to allow the development of a generalised approach in the sizing ofconstructed stormwater wetland that incorporates the complexity of the polluto-graph characteristics. A realistic approach to sizing stormwater wetlands will adoptthe mean event concentration as representative of the pollutant loading characteris-tics of a design event.

4. Adapting the k-C* model for stormwater wetlands

The design approach in determining the size of a constructed stormwater wetlandwould generally be directed at achieving a balance between the two objectives ofachieving the required hydrologic effectiveness and the effluent water quality target.

Table 2Pollutographs charachteristics–combined sewer system, Munich, Germany

Pollutograph characteristics Percentage of events

TSS BOD5 COD TOC KJN TP

Uniform 6.1%7.2% 0.0% 7.3% 2.9% 4.1%61.8% 57.3% 60.5%Early flush 65.7%68.8% 67.5%

Delayed flush 33.4%24.0% 34.3% 25.2% 35.3% 38.6%331223412332125Total number of events analysed


The effluent water quality target is set to reflect community standards, anddownstream ecosystem tolerance while the hydrologic effectiveness criterion sets thecompliance level in terms of the quantity of runoff meeting the water quality target.

The hydrologic effectiveness and the meteorological characteristics for the indi-vidual region can be related to the volume of the wetland according to Fig. 1 andEq. (5a) or (5b) can be used to determine the effluent quality for the given wetlandarea.

4.1. The equi6alent steady flow (Qesf)

An equivalent steady flow (Qesf) needs to be determined to enable Eq. (5a) or (5b)to be used in computing the expected effluent quality. This equivalent steady flowrate should reflect the hydraulic condition experienced by the pollutant within thewetland and is used in combination with the wetland area and the parameter k todetermine the pollutant removal rate within the wetland. Owing to the intermittentnature of inflow to the wetland, the detention period of pollutant entering thewetland varies according to the hydrograph and storage characteristics. The timedifference between the centroids of inflow and outflow hydrographs does notnecessarily represent the average detention period of the pollutant. This is especiallythe case for wetlands with a permanent pool and for situations where the volumeof the inflow hydrograph is significantly less than the volume of the wetland.

The computation of the mean pollutant detention period is further complicatedby the intermittent nature of inflow. For wetlands with a permanent pool, the waterquality of the initial stages of the outflow hydrograph represents the ambient waterquality of this permanent pool (Linforth et al., 1995). In such circumstances, thepollutant detention period is dependent on a number of factors including the ratioof the volume of inflow hydrograph, the storage volume of the permanent pool andthe dry weather period until the next storm event. Two general scenarios arepossible and they require a different approach towards computing the meanpollutant detention period, i.e.,1. If the volume of the inflow hydrograph is smaller than the permanent pool

volume, a significant portion of the inflow pollutant would be detained in thepermanent pool until the occurrence of the next event. Analysis of the sequenceof storm events using stochastic simulations will be necessary to compute thecombination of storm events and dry periods between events (Wong and Somes,1995);

2. If the volume of the permanent pool is small in comparison to the volume of theinflow hydrograph, the mean pollutant detention period may be computed bycalculating the time difference of the centroids of inflow and outflow hydro-graphs, but with the centroid of the outflow hydrograph adjusted for thewetland permanent pool volume deemed to have been discharged at the earlystages of the outflow hydrograph. This adjustment would have the effect ofshifting the centroid of the outflow hydrograph further away from the centroidof the inflow hydrograph.


It is evident that the pollutant detention period will vary according to the systemhydrology and the characteristics of the pollutograph in intermittently loadedwetlands. A representative pollutant detention period can only be studied in detailusing a continuous simulation approach. In the absence of a continuous simulationprocedure, an alternative means of approximating the representative pollutantdetention period needs to be developed. Under unsteady inflow conditions, theinflow hydrograph is often significantly attenuated in wetland with outlet hydraulicscontrolled by riser structures (i.e. vertical pipes with orifices along the length of thepipe) and the outflow hydrograph is characterised by a long duration of nearconstant rate of outflow. As a broad-based approach, the equivalent steady flow(Qesf) is best computed in the absence of continuous simulation as the ratio of thevolume of the wetland to the pollutant detention period (td):

Qesf=Vtd

(6)

The detention period, td, can vary with discharge. Typically, under a single orificeoutlet, the discharge condition is that of weir flow initially until the depth ofinundation exceeds the soffit level of the orifice at which point flow condition is thatof orifice flow. The form of the storage-discharge relationship is non-linear and thusthe detention period–discharge relationship can also be expected to be non-linear.Generally this is an undesirable form of detention period–discharge relationship asdetention periods during low discharges are often low owing to a more efficientdischarge characteristics at low levels of inundation. Riser outlets involve a numberof small orifices and experience with riser discharge characteristics indicates thatnear constant detention period for the full depth range of the wetland (i.e. a nearconstant ratio of storage volume to discharge) can be readily established byappropriate placement of orifices along the riser. Substituting Eq. (6) into Eq. (5a)and rearranging gives

td=hk

ln�Ci−C*

Co−C*�

(7)

4.2. Computation procedure

The computation procedure for adapting the k-C* model for designing con-structed stormwater wetlands will require the following information.1. The influent event mean concentrations, Ci.2. The background concentration, C*.3. The rate constant for the water quality parameter being treated, k.4. The effluent water quality target, Co.5. The compliance probability, i.e. the Hydrologic Effectiveness (c) of the wetland.Given this information, the computation steps for adapting the ‘k-C* model’ forstormwater wetlands can be summarised as follows:1. Compute the desired detention period necessary to meet the water quality target

for the influent event mean concentration Ci


2. Determine the required storage for the desired compliance probability using theHydrologic Effectiveness Curves

3. Compute the area required assuming average depth of 0.5 m above thepermanent pool level.

For each pollutant type, the procedure will derive to an area which achieves thebalance in meeting the water quality target and hydrologic effectiveness criteria.The appropriate area to be adopted for design is the largest area obtained amongstthose computed for the pollutants being considered. Appendix A presents a casestudy to demonstrate the application of this procedure for reduction of TSS andBOD.

5. Model parameters and research direction

The rate constants k and the background concentrations C* for the variouspollutants listed in Table 1 are based on steady flow conditions. Unsteady intermit-tent inflows are expected to have an influence on these values. This is a topic formuch research and will require field data from constructed stormwater wetlands.The comments and recommendations outlined in this section are made in theinterim to facilitate further discussion and research on how existing data can beused for stormwater wetland design. On-going research on constructed stormwaterwetlands is the only basis for advancing our understanding in these areas beyondthe current procedure.

5.1. Pollutants Affected by Physical Treatment Processes

Pollutants that are pre-dominantly affected by sedimentation include TSS andassociated attached metals and chemical compounds. Discussion below focusesmainly on TSS but may have relevance to other pollutants that are similarlyaffected by the sedimentation processes in the wetland.

5.1.1. Rate constant ktss-settling 6elocityWith sedimentation processes being the dominant mechanism for removal of

TSS, the settling velocity of suspended particles may be used as a measure of therate constant k. The distribution of particle settling velocities is related to thegrading, shape and density of the particles entering the wetland. Settling velocitiesmeasured in the laboratory can only be an indicator of the order of magnitude ofthe parameter k for TSS. Other factors such as resuspension due to non-ideal flowconditions can be expected to significantly reduce the effective settling velocities.Wetland vegetation can have the effect of increasing the magnitude of the rateconstant k as demonstrated by Lloyd (1997).

As indicated in Table 1, the range of k values is between 1000 and 10 000 m/yrand they correspond to particle sizes (diameter) of between 7 mm and 20 mm. Thissize range is typical of the material between the coarse silt to the fine sand range,and is appropriate for use in stormwater wetlands.


The following are some brief comments on the likely positive and negative effectson the rate constant as a result of unsteady intermittent inflows:� Experience with some data gathered from research projects undertaken by the

authors has indicated that the cyclic filling and draining of the wetland canfacilitate the adhesion of fine particles on vegetation surfaces leading to a higherkTSS value.

� Tracer studies as well as two dimensional hydrodynamic modelling by variousresearchers have found the flow hydrodynamics within the wetland during itsfilling and draining stages to be grossly two (or even three) dimensional. Theflow of water through densely vegetated sections of the wetland during thesephases is envisaged to have a positive effect on trapping suspended solids leadingto a higher kTSS value than applicable for steady flow systems.

� The vegetation cover reduces the potential for solid resuspension. Data fromstudies of TSS reduction in ponds and wetlands in Canberra, Australia (ACTAdministration Interim Planning Authority, 1990) have shown that a 20%increase in TSS removal can be gained by the introduction of macrophytes in awet detention system.

� The unsteady inflow conditions prevalent in stormwater wetlands would lead toa higher tendency for fine solids to be resuspended in these systems leading to alower kTSS value than applicable to steady flow systems. Solid resuspension isexpected to have the highest tendency near the inlet where the inflows have notbeen subjected to the full effect of storage attenuation. However, proper designof inlet structures as well as vegetation layout can significantly mitigate thiscondition.In the interim, it is suggested that the appropriate value of kTSS in stormwater

wetland be based on the settling velocity of the 50 percentile sediment grade withadjustments for increased effectiveness for wetlands with high vegetation density toreflect the experience from the Canberra study. This adjustment may be made bymultiplying kTSS by 1.2 if more than 50% of the wetland area is vegetated with asliding scale to 1.0 for vegetation density between 50% and zero.

5.1.2. Background ConcentrationBackground concentration of TSS for a given wetland is a reflection of the

characteristics of the substrata of the system and the characteristics of suspendedsolids generated from the catchment. Table 1 lists the equation for C* adopted byKadlec and Knight (1996) as being a linear function of the inflow concentration. Asingle measure using the inflow TSS concentration is thought to be insufficient inrelating the influence of hydraulic loading, particle size distribution, substrataconditions, biota growth etc on the background TSS concentration. The hydraulicloading and the size grading of the deposited sediment is envisaged to have a directinfluence on the amount of solids which can be resuspended and kept in suspensionin the water column. The dry period between events also have an influence on thestructure of settled particle as well as the development of algal solids.

It is envisaged that background concentrations of TSS will also be related tooutflow rate as flow velocities within the wetland is considered to be the primary


source of energy in the resuspension and maintenance of suspended particles in thewater column. It is possible for the relationship between background concentrationand flow rate to be derived from the field by regular monitoring of effluentconcentrations during low flow conditions. Furthermore, it is possible to conductsimple experiments involving the filling of a wetland (by closure of the outletstructure) and letting the water remain in the wetland for some extended period oftime. The wetland is then allowed to drain and TSS concentration determined at arange of outflow rates. These experiments could be carried out for differentantecedent conditions related to the duration of the dry period following the lastevent to account for the effect of this duration on the structure of settled particlesand algal growth in the unvegetated areas.

5.2. Pollutants affected by biological treatment processes

A category of pollutants that are influenced by a combination of physical,chemical and biological processes include such pollutants as BOD, COD, TN,TKN, NH4–N and PO4. The degree at which one type of process dominates theoverall treatment is dependent on various factors related to the characteristics ofthe pollutant, the chemical and biological state of the wetland etc. It is unlikely thatone form of treatment process will completely dominate the system. Discussionbelow is mainly directed to BOD but may have similar relevance to other pollutantsin this category.

5.2.1. Rate Constant kBOD

There is little known about how kBOD can be affected by unsteady intermittentinflows to the wetland. Intuitively, kBOD for stormwater wetlands ought to be higherthan corresponding values in wastewater wetlands for a number of plausiblereasons, including the following:1. A higher proportion of BOD in particulate form compared with soluble BOD is

expected in stormwater runoff and thus a significant amount of BOD reductionmay be associated with TSS reduction.

2. The opportunities for the ecosystem to recovery following a period of high BODloading during the inter-event periods.

3. The unsteady hydrodynamic within the wetland would have a higher DOreaeration potential thus able to satisfy BOD more rapidly.

5.2.2. Background ConcentrationRegression analysis of background BOD concentrations by Kadlec and Knight

(1996) found a correlation of this parameter with the influent BOD concentrationin wastewater wetlands. It is envisaged that intermittent inflow conditions instormwater wetlands could diminish the dependence of influent BOD concentrationand that the background concentration of stormwater wetlands would more reflectthe wetland characteristics including its vegetation type(s) and density, and soiltype. Kadlec and Knight (1996) found mean background BOD concentration inmarshes to be approximately 6 mg/l and data from a limited number of forested


wetlands found significantly lower background BOD concentration. It is envisagedthat the background BOD concentration in stormwater wetlands could be lowerthan that observed for marshes.

6. Conclusions

Understanding the performance of constructed wetlands in wastewater treatmenthas progressed to a stage where there is sufficient data to enable empirical modelsto be developed and their parameters calibrated. The two-parameter ‘k-C* model’,which models the rate of pollutant removal as an exponential function, is one suchmodel. Numerous field measurement of a range of pollutants have found this modelto satisfactorily represent pollutant concentration reduction along transects intreatment wetlands. The two parameters of the model, k and C*, are utilised toencompass the complex inter-relationship between physical, chemical and biologicalfactors which influence the performance of a wetland in reducing pollutant concen-trations. These parameters can be calibrated for individual sites from pilot studiesor during the early periods of the wetland operation and a good database ofparameter values are available for preliminary design of constructed wastewaterwetlands.

A possible framework for developing a basis for understanding and predictingthe performance of stormwater wetlands could take a similar strategy as that takenin wastewater wetlands. In the first instance, a broad means of predicting theperformance of stormwater wetlands can be achieved by developing empiricalrelationships or models that have a physical basis. The physical basis lies in theassociation between the parameters of the empirical model and physical, chemicaland biological processes in the wetland. Consideration is given to the adaptabilityof the ‘k-C* model’ for stormwater applications. A procedure has been formulatedwhich incorporates the use of the ‘k-C* model’ and the interaction between therequirements of wetland storage for inflow stochasticity and stormwater treatment.This procedure can be used to provide interim guidance on the proper sizing ofconstructed stormwater wetlands for long-term performance in stormwater treat-ment. Much research however needs to be conducted to understand the influence ofinflow stochasticity on the parameters of the ‘k-C* model’.

Appendix A. Case study: Melbourne wetlands

A wetland in Melbourne is proposed for treatment stormwater runoff from a 1km2 catchment area in the central business district.� Hydrology: The mean annual rainfall of Melbourne is 660 mm. It can be assumed

that the volumetric runoff coefficient for the Central Business District is 0.65 andthe Hydrologic Effectiveness Chart for this area is as shown in Fig 1.

� Water Quality: Two water quality parameters are to be treated, i.e. (1) BOD and(2) TSS. The maximum mean event concentrations of these parameters instormwater runoff from the catchment are 12 mg/l and 140 mg/l respectively.


� Wetland Performance Criteria: The long-term target water quality for BOD andTSS are 6 mg/l and 16.5 mg/l respectively, and these targets are to be met for90% of all stormwater runoff from the catchment.

� Model Parameters: For this case study, assume the following rate constant andbackground concentrations: BOD, kBOD=34 m/yr and C*BOD=4 mg/l; TSS,kTSS=1000 m/yr and C*TSS=15 mg/l

The three steps involved in the computation are as follows:1. Compute the desired detention period necessary to meet the water quality target

for the influent event mean concentration Ci, (Eq. (7)).2. Determine the required storage for the desired compliance probability using the

Hydrologic Effectiveness Curves (Fig. 1);3. Compute the area required assuming average depth of 0.5 m above the

permanent pool level.

TSS (kTSS=1000 m/yr; C*=15 mg/l)Steps BOD (kBOD=34 m/yr;C*=4 mg/l)

1td=

hk

ln�Ci−C*

Co−C*�

td td=hk

ln�Ci−C*

Co−C*�

td=0.5

1000ln�140−15

16.5−15�

=0.002y or 19 h; say 24 h=

0.534

ln�12−4

6−4�

=0.0204y

or 179 h; say 180 hFor c=90%, td=24 h; storage volume=For c=90%, td=180 h;22.0% of mean annual runoff volumestorage volume=4.3% of

mean annual runoff volume Vol=0.020×0.65×0.66×1 000 000Vol=0.043×0.65×1 000 000 =8580 m3

=18 447 m3

Assuming a depth of 0.5 m;Assuming a depth of 0.5 m;3

A=Volumedepth

=18 447

0.5A=

Volumedepth

=85800.5

=17 160 m3

=36 894 m3

The selected volume of the wetland wouldThe selected volume of thewetland would be approxi- be approximately 8600 m3 and the hydraulic

control would be designed to drain 90% ofmately 18 500 m3 and thehydraulic control would be the full storage volume over 24 h. The wet-

land area represents approximately 2.0% ofdesigned to drain 90% ofthe full storage volume over the catchment area or 3.1% of the equiva-

lent area.180 h. The wetland arearepresents approximately3.7% of the catchment areaor 5.7% of the equivalentimpervious area.


For this case study, BOD is the critical pollutant requiring a larger wetland to meetthe target effluent water quality criterion. The wetland area required is more thantwice that required for TSS treatment and for this size wetland, the expected meanevent outflow concentration of TSS can be calculated using Eq. (5b). In this case,the outflow concentration of TSS approaches the background concentration (C*) of15 mg/l.

References

Australian Capital Territory Administration Interim Planning Authority, 1990. Water Pollution ControlPond Design Guidelines, pp. 50.

Geiger, W.F., 1984. Mischwasserabfluss und dessen Beschaffenheit-ein Beitrag zur Kanalnetzplanung,Ph.D. thesis, Technical University of Munich, Publication No. 50.

Geiger, W.F., 1987. Prefilling of detention basins and its effect on overflows, Proceedings of the FourthInternational Conference on Urban Storm Drainage, XXII IAHR Congress, Lausanne, Switzerland.

Kadlec, R.H., 1996. Deterministic and stochastic aspects of constructed wetland performance anddesign, Proceedings of the Fifth International Conference on Wetland Systems for Water PollutionControl, Vienna, Austria, pp. X/I.1-X/I.8.

Kadlec, R.H., Knight, R.L., 1996. Treatment Wetlands, CRC Press, Boca Raton, FL., pp. 893.Linforth, S., Yu, S., Constandopolous, J., 1995. On the hydrodynamics of constructed wetlands,

Proceedings of the National Conference on Wetlands for Water Quality Control, Townsville,Australia, pp. 435–444.

Lloyd, S. D., 1997. Influence of Macrophytes on Sediment Deposition and Flow Pattern within aStormwater Pollution Control Wetland, MEngSc Thesis, Monash University, Victoria, Australia.

Somes, N.L., Wong, T.H.F, 1994. Design criteria for a trial wetland to improve the water quality ofrunoff from a small rural catchment, Proceedings of the International Hydrology and WaterResources Symposium of the Institution of Engineers, Australia, Adelaide, Australia, pp. 403–409.

Somes, N.L.G., Breen, P.E., Wong, T.H.F., 1996. Integrated hydrologic and botanical design ofstormwater control wetlands, Proceedings of the Fifth International Conference on Wetland Systemsfor Water Pollution Control, Vienna, Austria, vol 1, pp. III/4.1-III/4.8.

Wong, T.H.F., Somes, N.L.G., 1995. A stochastic approach to designing wetlands for stormwaterpollution control, IAWQ Water Sci. Technol., 32 (1), 145–151.

.

adaptation of wastewater surface flow wetland formulae for application in constructed stormwater...

Documents