Embed Size (px)
Citation preview
at SciVerse ScienceDirect
Journal of Environmental Management 119 (2013) 67e75
Contents lists available
Journal of Environmental Management
journal homepage: www.elsevier .com/locate/ jenvman
An integrated model of substrate clogging in vertical flow constructed wetlands
G.F. Hua a,b,*, L. Li b,c, Y.Q. Zhao d, W. Zhu e, J.Q. Shen f
aCollege of Water Conservancy and Hydroelectric Power, Hohai University, No. 1 Xikang Road, Nanjing 210098, PR Chinab State Key Laboratory of Hydrology-Water Resource and Hydraulic Engineering, Hohai University, No. 1 Xikang Road, Nanjing 210098, PR ChinacNational Centre for Groundwater Research and Training, School of Civil Engineering, The University of Queensland, St Lucia, SLD 4072, AustraliadCentre for Water Resources Research, School of Architecture, Landscape and Civil Engineering, University College Dublin, Belfield, Dublin 4, IrelandeCollege of Environment, Hohai University, No. 1 Xikang Road, Nanjing 210098, PR ChinafBusiness School, Hohai University, No. 1 Xikang Road, Nanjing 210098, PR China
a r t i c l e i n f o
Article history:Received 14 August 2012Received in revised form4 January 2013Accepted 27 January 2013Available online 1 March 2013
Keywords:Constructed wetland (CW)Substrate cloggingMicrobial biomassPlant rootsOrganic suspended solids
* Corresponding author. College of Water ConservaHohai University, No. 1 Xikang Road, Nanjing 210098
E-mail address: [email protected] (G.F. Hu
0301-4797/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.jenvman.2013.01.023
a b s t r a c t
This paper presents an integrated model of substrate clogging in a vertical flow constructed wetland(VFCW). The model simulates the reduction of pore space in the wetland substrate due to combinedinfluences of various physical, biogeochemical and plant-related processes. A series of experiments basedon laboratory-scale VFCWs were conducted to examine and measure key parameters related to cloggingof the wetland substrate during operation under different conditions. The model was then validatedusing data collected from the experiments. The results showed that the model was able to replicate theclogging phenomenon as observed in the experiments, in particular, the characteristic clogging time. Themodel also predicted well individual contributions to clogging by accumulated inert suspended solids,microbial biomass and plant root materials during the wetland operation. Although the validation wasbased on the laboratory data, the results indicated that the model describes well the processes under-lying the clogging and has the potential to become a tool for assessing the performance of prototype CWsin relation to clogging at both the design and operation stages.
� 2013 Elsevier Ltd. All rights reserved.
1. Introduction
Substrate clogging is a tangible risk for a subsurface-flow con-structed wetland (SFCW), affecting seriously the wetland’s functionin wastewater treatment. Clogging is mainly influenced by loadingrates of BOD, COD and suspended solids (SS) (Winter and Goetz,2003). During the wetland operation, SS and particulate organicmatters are removed from the treated water initially by means ofphysical processes such as sedimentation, entrapment andadsorption. Subsequently, the trapped SS and particulate organicmatters are hydrolyzed and degraded by aerobic and/or anaerobicprocesses (García et al., 2010). Dissolved organic matters can beadsorbed to granular media, plant roots and detritus, and oxidizedby resident microbial populations (Burgoon et al., 1995). Theremoval of these materials produces immobile solids that reduceand block the pore space. This in turnmay lead to dramatic changesof the flow condition in the system, creating problems such asshort-circuiting, reduction of hydraulic retention time, surface
ncy and Hydroelectric Power,, PR China.a).
All rights reserved.
ponding of wastewater, odors, presence of insects and after allconsiderable reduction in the wetland’s treatment efficiency(Platzer and Mauch, 1997). Under the worst-case scenarios, thelife span of the wetland treatment system may be significantlyshortened.
Although pore blockage is a common and pervasive problem inhorizontal- and vertical-flow treatment wetlands, very few quan-titative studies have been carried out to predict the rate at whichclogging occurs. A simple theoretical clogging model was devel-oped by Blazejewski andMurat-Blazejewska (1997) to calculate theclogging time based on a sandy CW system. In this model, a bio-solid density was introduced to convert the wastewater load toporosity reduction. Langergraber et al. (2003) and Zhao et al. (2004)presented an approach for estimating the clogging time based onthe concept of available void space in the substrate. WhileLangergraber et al. (2003) considered the degradable fraction ofaccumulated matters, Zhao et al. (2004) focused on acceleratedsolids mineralization during operation-break periods. Similarly, inour previous study, a conceptual model has been developed byusing the parameter of influent SS concentration to estimate theclogging time (Hua et al., 2010). However, all these models did nottake into account the biofilm growth, or the influence of vegetationand its contribution to the increase of recalcitrant detritus.
G.F. Hua et al. / Journal of Environmental Management 119 (2013) 67e7568
On the other hand, Rousseau (2005) developed a mechanisticmodel that estimates the reduction of pore volume in CWs asa function of time. The model is able to predict the porosity changeand thus can be a useful tool for studying the clogging phenome-non. However, this model contains 26 state variables associatedwith 26 mass balance equations, including in total 118 model pa-rameters. With so many parameters involved, it is extremely dif-ficult to calibrate the model in order to determine proper values ofinput parameters, especially those that are not measurable orcannot be measured easily. This would lead to great model uncer-tainty in applications (García et al., 2007). Moreover, this model didnot include the macrophyte root biomass in the calculation of porevolume reduction. In theory, the clogging time estimation shouldconsider the detached biofilm and plant detritus in the CWs.
Giraldi et al. (2010) developed a one-dimensional finite elementanalysis (FEA) model (named FITOVERT) for VF treatment wetlandsto evaluate clogging due to biofilm and particulate matter accu-mulation. The coupling between clogging and hydraulics wasachieved by relating the accumulation of clogging matter to loss ofmedia porosity. Only the hydraulic module of the model was cali-brated and validated. Furthermore, the contribution of the plantroot’s biomass to clogging was not considered either.
The aim of this study is to develop a clogging model to simulatekey physical (such as physical filtration), biogeochemical and plant-related processes that are responsible for the pore space reductionduring the operation of VFCWs. This model is based on a relativelysmall set of parameters that describe adequately the characteristicsof wastewater (e.g., SS and COD concentration etc.), wetland me-dium/substrate and plant. In parallel to the model development,experiments were conducted within laboratory-scale VFCWs. Datacollected from these experimental wetlands were used to calibrateand validate the model.
2. Model development
2.1. Background
It is well recognized that the treatment of wastewater in CWsresults from various physical, biogeochemical and plant-relatedprocesses. SS in the wastewater is mainly removed near the inletof the wetland, typically within 1/3 of the total depth from thesurface, under normal operating conditions (Wynn and Liehr,2001). Only at high flow rates, wash-out of solids proportional tothe flow rate occurs (Rousseau, 2005). Although detachment ofbiofilm is a commonly acknowledged process, it is assumed thatsloughed parts of the biofilm are retained and metabolized withinthe pores, unless they are washed out by a peak flow.
Aerobic and anoxic microbial carbon conversion processes inCWs are often simulated based on the Activated Sludge Model(ASM1; Henze et al., 2000) and Constructed Wetland Model 1(CWM1 Langergraber et al., 2009). These modeling approacheswere also adopted in the present study with improvements of theoriginal models made to better represent CWs by incorporating theeffects of vegetation. Following the method of Wynn and Liehr(2001), the plant growth and decay model was kept simple, withno explicit consideration of individual factors such as nutrientavailability, air temperature, solar radiation and water level.
Clogging was evaluated based on pore volume reduction in thepresent model. Three contributors to pore volume reduction were:(1) biofilm due to bacteria growth, (2) solids retained and (3) plantdetritus. The model computed the masses of these substances,which were then converted to volumes according to the massdensities to determine the pore volume reduction. If the amount ofaccumulated matters became so high that there was virtually noavailable void space inside the pores for further settlement of solid
materials, the clogging occurred. At this point of time, the hydraulicconductivity of the CW substrate decreased significantly witha near-zero substrate porosity. The period of the CWoperation up tothis point of clogging occurrence was defined as the clogging time.
2.2. Model setup
The model development was based on the mass balance ofinvolved materials in the CW, following the principle given below:
dM TSdt
¼ Fin � Fout þ S (1)
M_TS is the total solid mass (g) in the wetland system (thebiomass of plant roots considered separately); t is the time (d); theFin and Fout are the influx and efflux (g/d), respectively; S is thesource/sink term (g/d) due to reaction/conversion.
As shown in Fig. 1, the source/sink terms in Equation (1) dependon the following processes and factors: (a) total inert suspendedsolids are absorbed and retained inside the substrate; (b) bio-degradable organic matters are converted into biosolids; (c) bio-mass residues are transformed into inert solids through microbialendogenous respiration. Note that inert solids can also be derivedfrom plant roots considered separately.
Calculations of the flux and source/sink terms are discussed indetail below:
(I) The influx of inert particulate matter (Fin_IS, g/d) associatedwith the influent to the CW, including organic and/or inor-ganic substance (MIS), can be expressed as:
Fin IS ¼ Qin � Cin � ð1� fv þ fv � fnvÞ (2)
where Qin is the flow rate of the influent (m3/d); Cin is the influentconcentration of TSS (g/m3); fv is the proportion of organic matterin the TSS (�); and fnv is the proportion of the inert (non-bio-degradable) matter in the organic TSS (�).
(II) The source of biosolids production (SBS, g/d) due to the con-version of biodegradable organic matter can be expressed as:
SBS ¼ Qin � CBOD � YH1þ K � q
(3)
dwhere CBOD is the influent BOD (g/m3) concentration; Kd is theheterotrophic microbial endogenous decay coefficient (d�1); YH isthe observed yield for heterotrophic biomass (�) and q is the meanresidence time (d) of biosolids in the CW.
(III) The source of inert matter production from conversion ofbiomass residues (SIS, g/d) through microbial endogenousrespiration can be described as:
SIS ¼ fp � Kd � q� SBS (4)
where fp is the fraction of microbial biomass converted to inertmatter (�).
Some fractions of inert solids and biomass solids would bewashed off and discharged through the effluent. Thus, the massbalance equations for inert solids (M_IS, g) and biomass solids(M_BS, g) in the system are:
dM ISdt
¼ Fin IS � Fout IS þ SIS (5)
Fig. 1. Schematic diagram of the CW system including the processes involved.
G.F. Hua et al. / Journal of Environmental Management 119 (2013) 67e75 69
dM BSdt
¼ Fin BS � Fout BS þ SBS (6)
The accumulation of total solid mass (M_TS) in the pore space ofthe substrate is then given by:
dM TSdt
¼ dM ISdt
þ dM BSdt
(7)
(IV) The effluxes of both the inert solids and biomass solids areassumed to be proportional to their respective influxes, i.e.,
Fout ¼ l� Fin (8)
where l is a ratio varying between 0 and 1. Although Equation (8)overly simplifies the involved processes such as biofilm detach-ment, the assumption is consistent with the removal rate conceptused in designing the CW systems, and shown to be reasonablethrough comparison between model predictions and experimentaldata presented in Section 4.
Combining Equations (5)e(8) leads to the following governingequation of the total solid mass accumulated in the substrate:
dM TSdt
¼ �1� l
�� �Fin IS þ Fin BS
�þ SBS þ SIS (9)
Equation (9) can be solved to determine M_TS using the Eulermethod, i.e.,
M TSnþ1 ¼ M TSn þhð1� lÞ �
�Fnin IS þ Fnin BS
�þ SnBS þ SnIS
i�Dt
(10)
where Dt is the size of the time step (set to a day in the currentstudy) and the superscript indicates the particular time step.
The roots of wetland plants can also cause clogging of thepore space in CWs. To determine the increasing volume of theplant roots associated with plant growth, a plant dynamicsmodel was developed based on plant biomass (Mplant, g) in both
the living (MplantL, g) and dead (MplantD, g) forms, i.e.,Mplant ¼ MplantL þ MplantD. During the growth season, the plantgrowth, assisted by large amounts of nitrate and/or ammoniumavailable in the wastewater, leads to increase of the living plantbiomass. As the growth season passes, the plant decay dominates,with the living biomass converted into dead biomass (Wynn andLiehr, 2001). Thus, the plant growth can be expressed as:
�1�Growth season :
dMplantL
dt¼ MplantL �
�fNH þ fNO � bp
�(11a)
�2�Non� growth season :
dMplantL
dt¼ �MplantL � bp
(11b)
where bp (day�1) is the plant decay coefficient for living plantmaterial (note that plant decay occurs during both the growth andnon-growth seasons); and fNH (day�1) and fN (day�1) are the plantgrowth rates depending on ammonium and nitrate concentrations,respectively, i.e.,
fNH ¼ kpl ��
CNHKPNH þ CNH
�(12)
fNO ¼ kpl ��
CNOKPNO þ CNO
���
KPNH
KPNH þ CNH
�(13)
where kpl (day�1) is the relative plant growth rate; KPNO (gNO3eN/m3) is the nitrate half-saturation coefficient for plant growth; KPNH
(gNH4eN/m3) is the ammonium half-saturation coefficient forplant growth; and CNH (g/m3) and CNO (g/m3) are ammonia andnitrate nitrogen concentrations of the influent, respectively.
As a result of the plant decay, the dead plant biomass (MplantD) isproduced and subsequently subjected to physical degradationprocesses due to, for example, wind action and invertebrate con-sumption; thus
G.F. Hua et al. / Journal of Environmental Management 119 (2013) 67e7570
dMplantD
dt¼
�MplantL � bp �MplantD � kdegradation
�(14)
where kdegradation (day�1) is the first-order decay rate constantrepresenting the loss of biomass from the CW system due tophysical degradation processes mentioned above.
Combining Equations (11) And (14) yields the dynamic equationfor the total plant biomass, i.e.,
�1�Growth season :
dMplant
dt¼ MplantL � ðfNH þ fNOÞ �MplantD
� kdegradation(15a)
�2�Non� growth season :
dMplant
dt¼ �MplantD � kdegradation
(15b)
Again, Equation (15) can be solved numerically using the Eulermethod to determine the total plant biomass in the CW. The bio-mass of the plant root (MplantR), the underground portion of theplant mass, can then computed as follows (Mander et al., 2008;Collier et al., 2010),
MplantR ¼ Mplant � h (16)
where h is the root shoot ratio (�).The total volume associated with the total solids and plant roots
biomass can be determined based on the densities of these mate-rials, i.e.,
VT ¼ M TSrs � ð1� uÞ þ
MplantR
rplantR(17)
where rs and u are the density (g/cm3) and moisture content (%) oftotal solids, respectively and rplantR is the density (g/cm3) of theplant roots. Clogging can then be evaluated: if VT is so high thatthere is virtually no available void space inside the substrate forfurther settlement of TS and plant roots, clogging occurs. The timeof operation required to reach the point of clogging is defined as the“clogging time” (tc). This critical condition is given by
VT � 3� hc � A (18)
where 3� hc � A is the total pore volume in the wetland substratewith 3being the substrate porosity, A wetland surface area (cm2)and hc the depth of filtration/clogging layer (cm).
3. Laboratory experiments
3.1. Experimental wetland setup and operation
Six identical laboratory-scale wetland beds (made of perspexcolumns of 100 cm in height and 15 cm in diameter as shown in
Table 1Operational conditions of the six beds used in the laboratory experiments.
Influent Polluted river water
Bed 1 Bed 2
Planted Unplanted
COD (mg/L) 100 � 11Organic loading rates (g COD/m2 d) 50 � 5.5TSS (mg/L) 80 � 15TN (mg/L) 10.24 � 3.32TP (mg/L) 1.98 � 0.58
Fig. 1) were used in the study. The beds were filled with the samecoarse sands of d10 ¼ 0.12 mm to the same depth of 75 cm. Amongthe six beds, Bed 1, Bed 3 and Bed 5 were planted with Typhaangustifolia, which has an extensive root system, while Bed 2, Bed 4and Bed 6 remained unplanted for the purpose of investigating therole of plant in the clogging process. Prior to the experiments, theplants were allowed to develop during a 3-month start-up periodwhen colonies of microorganisms also started to form. The diam-eter of plant in a cluster density corresponded to that of the modelwetland to simulate the actual clogging condition. The outsidesurface of all the perspex wetland columns including the outletswas painted black to shield the columns from light to prevent algalgrowth.
Wastewater was prepared by adding to tap water the starch asthe source of slowly biodegradable suspended organic solids (Minoet al., 1995). Subsequently, (NH4)2SO4, CO(NH2)2 and K2HPO4 wereadded as major nutrients to the artificial wastewater to control theconcentrations of TN and TP at about 10 mg/L and 2 mg/L,respectively. The six CWs beds were operated in parallel withdownward flow and continuously loaded at a hydraulic loading rateof 0.5 m3/m2 d from November 2008 to September 2009. Theoperational conditions of the six beds are summarized in Table 1.
The chemical concentrations in the experiments were set ac-cording to the average concentrations of three types of wastewateras polluted river water (Wu et al., 2007), storm runoff (Chu et al.,2008) and municipal wastewater (Yuan et al., 2004) in EasternChina. The balance between the organic load and oxygenationrenewal within the substrate is of great importance to the wetlandperformance (Platzer and Mauch, 1997; Kayser and Kunst, 2005).Thus the effects of different organic loads were also investigated.
3.2. Measurements
The pore volumes of the wetland beds were measured bydraining the preset water saturated bed. The measured drainagevolumes of each layer (via outlet) along the height of the beds wereused to estimate the pore volumes (Hua et al., 2010). This was doneevery 15 days until ponding clearly occurred on the bed surface.
TSS, COD, TN and TP were analyzed using standard methods asdescribed in Water and Wastewater Monitoring and AnalysisMethods (2002). The concentrations of TN, TP, COD and TSS of theinfluent were measured every day while the concentrations of CODand TSS of effluent weremeasured every 7 days. In addition, the DOof pore water was measured with the iodometric method com-monly known as the Winkler titration as described in Water andWastewater Monitoring and Analysis Methods (2002). The first twosteps are as follows:
(1) Fill a 50 mL BOD bottle with a sample from the sample con-nection of the wetland column in situ. Note that the bottleshould contain no air bubbles.
(2) Immediately, add 1 mL of manganous sulfate and 1 mL ofalkali-iodide-azide with a pipette into the subsurface of the
Storm runoff Municipal wastewater
Bed 3 Bed 4 Bed 5 Bed 6
Planted Unplanted Planted Unplanted
203 � 17 398 � 14101 � 8.5 199 � 7.0152 � 4 275 � 14
10.67 � 2.28 10.09 � 4.082.45 � 0.47 2.69 � 0.98
G.F. Hua et al. / Journal of Environmental Management 119 (2013) 67e75 71
sample. Then stopper and invert the bottle several times tomix. Thus, the DO level of the sample can be fixed to allow timefor transporting the sample to a laboratory for completion ofthe analysis. The sample should be kept cool and protectedfrom sunlight until the analysis is completed according to theremaining steps of the iodometric method.
Measurements of the oxygen uptake rate (OUR) were also madeto determine the yield coefficient (YH) and decay rate (Kd) of thebiodegradation according to themethod of ASM1 (Stasinakisa et al.,2002). The microbial samples for the OUR measurement weretaken from the clogging layer (on the surface) of the wetland beds.To examine the correspondence of the microbe growth with theclogging process, YH and Kd were measured on the 15th, 75th and120th operation day to represent the conditions at the early, stableand later clogging stage, respectively.
The density of plant root was determined according to the root’squantity and its volume. At the end of the experiments, the fullydeveloped plant roots were taken out from the systems and putinto the beaker filled with distilled water. The root volume wasobtained using a drainage method based on the displaced watervolume.
4. Results
4.1. Parameters estimation
Table 3 (see Supplementary material) summarizes the values ofall the model parameters used in the simulations. The parametersincludephysical, operational,microbial and plant-related quantities.
Fig. 2 shows the variation of the measured YH values at differentoperation stages. YH appeared to increase initially and then declineslightly with the operation time. This trend is particularly evidentin the planted bed. While the initial increasing phase might belinked to adaptability of the microbial community to the CW con-dition and hence increased growth rate, the declining phase couldbe due to increasing inhibition effects of clogging on microbes asa result of reduced substrate aeration.
Measured YH values of the planted beds were found to be higherthan those of unplanted beds. At the stable operation stage (75days), a maximum difference of 0.1 g/g COD existed between the YHvalues of the planted beds (with the highest value of 0.64 g/g COD)and unplanted beds (with the lowest YH value of 0.54 g/g COD). Thismight be due to plant roots, which provided large surface areas formicroorganism attachment, leading to improvement of the degra-dation rate of the organic matters and promoting the biofilm
Operation time (days)
YH (
g/g
CO
D)
0.50
0.55
0.60
0.65
0.70
0.75
0 30 60 90 120 150
Bed1 Bed2Bed3 Bed4Bed5 Bed6
Fig. 2. Variation of YH at different operation stages.
growth. In addition, the bacteria populations in the planted andunplanted beds could be different; a mutualistic relationship mightexist between the microbes and plants in the planted beds.
The measured Kd values show a decreasing trend with theoperation time in both planted and unplanted beds (Fig. 3). Thismight be related to the reduction of dissolved oxygen (DO) due toclogging. The value of DO was found through measurements todecrease from 4.9 mg/L to 0.4 mg/L in Bed 1; and similar reductionswere also observed in the other five beds. The average Kd values onthe 15th, 75th and 120th operation day were 0.11 day�1, 0.07 day�1
and 0.04 day�1, respectively. It should be pointed out that in gen-eral, Kd is a constant, but can change in the studied CW system asthe microbial population evolves under increasing DO stress due toclogging.
Overall, the difference in the Kd value between beds subjected todifferent organic loads was small, for example, Kd ¼ 0.12 day�1 inBed 1 compared with and Kd ¼ 0.13 day�1 in Bed 3. This suggeststhat the effect of organic load on the Kd value was modest, indi-cating sufficient carbon sources for the respiration of aerobic mi-crobes in both beds. It is interesting to note that the Kd values of theplanted beds were higher than those of unplanted beds especiallyunder the low organic load condition. This is likely to be linked tothe effects of rhizosphere aeration.
No data could be found for the parameter of the biosolids meanresidence time in constructed wetlands with no exclusion processof the biomass sludge. Noting that this value in activated sludgesystems with extensive aeration is typically 25 days (Henze et al.,2008) and it could be higher in this study, we assumed a meanresidence time of 30 days. A sensitivity analysis was then conductedto assess the uncertainty of modeling results associated with thisparameter.
4.2. Sensitivity analysis
The sensitivity analysis was conducted based on the relativesensitivity index to examine changes in model output (X) due tochanges with a model parameter (Param),
Sx ¼ DX=XDParam=Param
(19)
10% of relative parameter value changes were considered. Theresults revealed that the sizes of the wetland bed, porosity and d10value of the substrate sediments were the most influential pa-rameters in affecting the clogging of the CW system, followed bythe daily flow rate and TSS concentration of the influent. Parame-ters controlling the plant growth have relatively small effects on the
Operation time (days)
Kd (d
-1)
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0 30 60 90 120 150
Bed1 Bed2Bed3 Bed4Bed5 Bed6
Fig. 3. Variation of Kd at different operation stages.
Clo
ggin
g tim
e (d
ays)
Operation time (days)
0
50
100
150
200
250
300
Bed1 Bed2 Bed3 Bed4 Bed5 Bed6
experimental
model prediction
Fig. 4. Measured and predicted clogging times.
G.F. Hua et al. / Journal of Environmental Management 119 (2013) 67e7572
simulated clogging time. It should be pointed out that although thesizes of the wetland bed appeared to be the most influentialparameter, the actual influence was due to the SS/organic load(Ns ¼ Qin � Cin/pD2, g (SS/COD)/m2 d), which depended on thediameter of the wetland column. All the results are summarized inTable 2.
The results concerning two particular parameters are furtherdiscussed in the following. First, since no data could be found forthe biosolids mean residence time, an approximate value of 30 dayswas assumed in the initial simulation. This residence timewas thenchanged to 3, 15, 25, 50, 100 and 200 days in the simulations for thesensitivity analysis. The corresponding predictions of the cloggingtime were 385, 390, 391, 393, 394, 394 days, respectively. Thissuggests that the biosolids mean residence time was not a keyparameter in affecting the simulated clogging time. Secondly, thevalue of l (ratio of efflux to influx) (Table 3, i.e., Supplementarymaterial) was from a previous study based on gravel beds(Rousseau, 2005). Further simulations were conducted with thefraction coefficient set to 0.14, 0.12, 0.1, 0.08, 0.06 and 0.04. Thepredicted clogging times were 389, 372, 342, 312, 287 and 251 days,respectively. Thus, the model output of the clogging time appearedto bemore sensitive to l but the variations of the predicted cloggingtime remained to be relatively small.
4.3. Measured and predicted clogging times
The measured clogging times of Bed 1 to Bed 6 are 258, 261, 198,215, 125 and 135 days, respectively. These times were determinedwhen the measured drainable pore volumes of the six beds werezero and the ponding condition was clearly evident on the bedsurface. The measured and predicted clogging times were found toagree reasonably well with each other (Fig. 4). This was confirmedby the t-test for significant differences between the model pre-dictions and experimental data, which gave a two-tailed p value of0.023, indicating insignificant difference at the level of a ¼ 0.01.
It is interesting to note that while the model predicted slightlylonger clogging times for the planted beds than those of unplantedbeds, the experimental results showed the opposite trend. The t-test for significant plant effect on substrate clogging under thesame influent condition (i.e., between the pairs of Bed 1 and Bed 2,Bed 3 and Bed 4, and Bed 5 and Bed 6) gave a two-tailed p valuesvalue of 0.605, 0.749 and 0.917, respectively, indicating insignificantplant effect at the level of a ¼ 0.05. During the experiments, theplant itself played aminor role in the clogging process. For example,in Bed 1 the underground biomass (plant roots) only increased by164.98 g/m2 over the 300 days’ growth period and occupied 1%e2%of the pore space in the clogging layer.
4.4. Pore space reduction
Fig. 5 shows measured and predicted pore space reduction. Itcan be seen that accumulation of the inert matter, microbial bio-mass and plant roots in the CW had a highly negative impact on thesystem’s porosity. In Bed 1, about 40% of the pore space in thesubstratewas filled within 220 days’ operation; and this percentagereached around 60% after another 70 days of operation. Similar
Table 2Results of sensitivity analyses.
Parameters Ns d10 n Qin Cin rSS fv u
Sx 0.51 0.51 0.46 0.48 0.48 0.46 0.11 0.12
Parameters fnv CNH CNO Mplant Kd YH l
Sx 0.48 0.00 0.00 0.03 0.09 0.14 0.48
trends occurred in the other beds but with different rates of porespace reduction as shown in Fig. 6. The pore space was filled morequickly in the beds subjected to higher COD loading.
Measured and simulated reductions of the total pore volumesagreed reasonably well, with the former slightly higher than thelatter. In all cases, the microbial biomass and plant roots occupiedrelatively small proportions of the pore space (6%e8%) comparedwith that taken by the inert matters (Fig. 7).
5. Discussions and concluding remarks
Further simulations were conducted to examine how the clog-ging time varies with the hydraulic and COD loading rates. Undera hydraulic loading of 0.5 m3/m2 d with a COD concentration of400 mg/L, the wetland with the substrate of coarse sand is pre-dicted to operate for 155 days before clogging occurs (fnv ¼ 0.8).Reducing the COD concentration to 100 mg/L increases theexpected period of operation to 292 days. In the case of hydraulicloading of 0.1 m3/m2 d and influent COD concentration of 50 mg/L,the wetland operation period is further extended to 1511 days.
While it is not a straightforward matter to compare the cloggingtimes from this studywith others in the literature since the cause ofclogging may differ from case to case, the clogging times measuredin the experiments and predicted by the current model appeared toagree with those reported previously. For example, Behrends et al.(2007) showed that about 30%e68% of the pore space was filledwith recalcitrant organic sediments in a VFCW system after 28months of operation, under a loading condition of 18.3 g COD/m2 d(0.5 m3/m2 d hydraulic load and 36.6 mg/L COD concentration ininfluent).
Themodel proposed in this study incorporates the contributionsof inert SS matters, microbial biomass and plant roots to the porespace reduction. Based on the systems examined in the laboratoryexperiments, biomass growth plays a minor role in controlling thewetland clogging compared to that of inert solid matters. Althoughthe influence of the organic load is an important factor leading tomassive growth of bacteria in the form of biofilm, especially froma long term point of view (Kim et al., 2010), the obvious accumu-lation of suspended solids in the top layer of the wetland substrateseems to dominate the clogging process, a result consistent withthe findings of Langergraber et al. (2003), Zhao et al. (2009) andGiraldi et al. (2010).
This study also showed that the wetland plant did not contrib-ute much to the pore space reduction, which is in alignment with
0
10
20
30
40
50
60
70
80
90
100
0 50 100 150 200 250 300
predicted total porosity reduction
predicted porosity reduction by microbial biomass
predicted porosity reduction by inert matter
predicted porosity reduction by plant roots
measured total porosity reduction
0102030405060708090
100
0 50 100 150 200 250
010
2030
4050
6070
8090
100
0 20 40 60 80 100 120 140 160
Pore
spa
ce r
educ
tion
(%)
Bed1
Operation time (days)
Bed3
Bed5
Pore
spa
ce r
educ
tion
(%)
Pore
spa
ce r
educ
tion
(%)
Fig. 5. Measured and predicted pore space reduction.
0
10
20
30
40
50
60
70
80
90
100
0 50 100 150 200 250 300 350
Bed1Bed2Bed3Bed4Bed5Bed6
Operation time (days)
Pore
spa
ce r
educ
tion
(%)
Fig. 6. Total porosity reduction.Fig. 7. Porosity reduction percentages due to inert matter, microbial biomass and plantroots.
G.F. Hua et al. / Journal of Environmental Management 119 (2013) 67e75 73
G.F. Hua et al. / Journal of Environmental Management 119 (2013) 67e7574
Langergraber et al.’s (2003) previous finding that the roots blockedonly a small fraction of the pores. The role of plants in SFCWsclogging remains unclear and debatable. It has been suggested thatplant can benefit thewetland bymaking the substratemore porous.For example, Fisher (1990) observed that pant root expansion in anAustralian field-scale horizontal subsurface flow treatment wet-land (HSSF TW) increased the depth of the reactor in the inlet zoneby 60 mm over three years, which would serve to counterbalancethe clogging effect. On the other hand, plant roots have also beenreported to have a detrimental effect on the CW system clogging.While studying dairy wastewater treatment HSSF TWs in NewZealand, Tanner (1994) showed the potential of clogging due toplant detritus. Moreover, Tanner and Sukias (1995) comparedorganic matter accumulation rates in planted and unplanted beds,and found that the accumulation rates were higher in the plantedsystems. Above the ground, up to half of the standing plant biomass(stems) may be dead, but remain relatively stable due to thepresence of esters in the plant tissue, which helps the plant to resistdegradation (Tanner et al., 1998). In addition, it is believed that thevertical gradient in plant establishment has a negative impact onthe hydraulic efficiency of CWs. Knowles et al. (2010) performedinternal hydrodynamic visualizations using dissolved tracers andreported observations of preferential flow along the bottom of thebed. Pedescoll et al. (2011) stated that macrophytes (reed) greatlycontributed to clogging; after three years of operation, rootsaccumulated in planted wetlands amounted to 35%e70% of thetotal accumulated solids. These different trends and behaviors werelikely to be due to different types of plants and roots, and thedepths of the deposition layer. It is interesting to note that while themodel predicted slightly longer clogging times for the planted bedsthan those of unplanted beds, the experimental results showed theopposite trend. This might be because in our model we only con-sidered the pore volume that the plant root occupied in thewetlandsystem to calculate the predicted clogging time and neglected theeffect of the plant root on the hydraulic characteristics and micro-bial community which may also affect the clogging process.
To speed up the clogging, a relatively high and continuous hy-draulic loading was adopted in this study. This mainly affected thedeposition and filtration of incoming suspended solids, especiallynear the surface. However, even under the high hydraulic load, thesuspended sediment concentration in the effluent remained verylow (measured SS concentrations of the effluent in the six bedswere all close to zero), which implied that the clogging processoperated similarly under different hydraulic loading conditions.Therefore, it is still reasonable to extrapolate the results obtained inthis study to more realistic systems. In addition, compared with theintermittent flow condition, the continuous flow operation modewould make the wetland system anaerobic, which would affect thebiofilm’s structure. Thus, we should note that the models byLangergraber et al. (2003) and Platzer and Mauch (1997) have beendeveloped for intermittently loaded and not continuously loadedVFCWs. One way to increase the life span of the beds is to haveintermittently instead of continuously loaded beds.
Substrate clogging in constructed wetlands is a complicatedphenomenon influenced by physical, biological and plant-relatedprocesses. In this study, an integrated approach has been pro-posed to investigate clogging by considering trapped solids, biofilmand plant roots accumulated in the CW system. The developedmodel has the ability to make a quantitative assessment of theclogging behavior and operation characteristic of the constructedwetland system. Based on the model, sensitivity analyses can beconducted to identify key parameters that affect the clogging.Further model refinement and calibration of important model pa-rameters are needed to enhance the model’s reliability in the ap-plications to assist designs of optimal CW systems.
Acknowledgments
The authors acknowledge the financial support for this workfrom the National Basic Research Program of China (Grant No.2012CB417005), National Higher-education Institution GeneralResearch and Development Funding (2012B00714), and NaturalScience Fund Project in Jiangsu Province (BE2011809).
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jenvman.2013.01.023.
References
Behrends, L.L., Bailey, E., Jansen, P., Houke, L., Simth, S., 2007. Integrated constructedwetland systems: design, operation, and performance of low-cost decentralizedwastewater treatment systems. Water Sci. Technol. 55 (7), 155e161.
Blazejewski, R., Murat-Blazejewska, S., 1997. Soil clogging phenomena in con-structed wetlands with subsurface flow. Water Sci. Technol. 35 (5), 183e188.
Burgoon, P.S., Reddy, K.R., DeBusk, T.A., 1995. Performance of subsurface flowwetlands with batchload and continuous-flow conditions. Water Environ. Res.67 (5), 855e862.
Chu, J.Y., Li, W., Li, W.B., 2008. Application of hydrocyclone separator to pollutioncontrol in storm runoff. Drain. Irrit. Mach. 26 (4), 57e60.
Collier, M.H., Keane, B., Rogstad, S.H., 2010. Productivity differences between dan-delion (Taraxacum officinale; Asteraceae) clones from pollution impacted versusnon-impacted soils. Plant Soil 329 (1e2), 173e183.
Fisher, P.J., 1990. Hydraulic characteristics of constructed wetlands at Richmond,New South Wales, Australia. In: Cooper, P.F., Findlater, B.C. (Eds.), ConstructedWetlands in Water Pollution Control. Pergamon Press, Oxford, United Kingdom,pp. 21e32.
García, J., Rousseau, D.P.L., Caselles-Osorio, A., Story, A., Pauw, N.D.,Vanrolleghem, P., 2007. Impact of prior physico-chemical treatment on theclogging process of subsurface flow constructed wetlands: model-based eval-uation. Water Air Soil Pollut. 185, 101e109.
García, J., Rousseau, D.P.L., Morató, J., Lesage, E., Matamoros, V., Bayona, M.J., 2010.Contaminant removal processes in subsurface-flow constructed wetlands:a review. Crit. Rev. Environ. Sci. Technol. 40, 561e661.
Giraldi, D., De Michieli Vitturi, M., Iannelli, R., 2010. FITOVERT: a dynamic numericalmodel of subsurface vertical flow constructed wetlands. Environ. Model. Softw.25 (5), 633e640.
Henze, M., Gujer, W., Mino, T., Van Loosdrecht, M., 2000. Activated Sludge ModelsASM1, ASM2, ASM2d and ASM3. Scientific and Technical Report No. 9. IWAPublishing, London, UK.
Henze, M., Van Loosdrecht, M.C.M., Ekama, G.A., Brdjanovic, D., 2008. BiologicalWastewater Treatment Principles, Modelling and Design. IWA Publishing,London, UK, pp. 53e86.
Hua, G.F., Zhu, W., Zhang, Y.H., 2010. A conceptual approach based on suspendedsolids to estimate clogging time in constructed wetlands. J. Environ. Sci. HealthPart A 45, 1519e1525.
Kayser, K., Kunst, S., 2005. Processes in vertical-flow reed beds: nitrification, oxygentransfer and soil clogging. Water Sci. Technol. 51 (9), 177e184.
Kim, J.W., Choi, H., Pachepsky, Y.A., 2010. Biofilm morphology as related to theporous media clogging. Water Res. 44 (4), 1193e1201.
Knowles, P.R., Griffin, P., Davies, P.A., 2010. Complementary methods to investigatethe development of clogging within a horizontal sub-surface flow tertiarytreatment wetland. Water Res. 44, 320e330.
Langergraber, G., Haberl, R., Laber, J., Pressl, A., 2003. Evaluation of substrate clog-ging processes in vertical flow constructed wetlands. Water Sci. Technol. 48 (5),25e34.
Langergraber, G., Rousseau, D.P.L., García, J., Mena, J., 2009. CWM1: a general modelto describe biokinetic processes in subsurface flow constructed wetlands. WaterSci. Technol. 59 (9), 1687e1697.
Mander, Ü., Lõhmus, K., Teiter, S., Mauring, T., Nurk, K., Augustin, J., 2008. Gaseousfluxes in the nitrogen and carbon budgets of subsurface flow constructedwetlands. Sci. Total Environ. 404 (2e3), 343e353.
Mino, T., San, D., Matsuo, T., 1995. Estimation of the rate of slowly biodegradableCOD (SBCOD) hydrolysis under anaerobic, anoxic and aerobic conditions byexperiments using starch as model substrate. Water Sci. Technol. 31 (2), 95e103.
Pedescoll, A., Corze, A., Álvarez, E., García, J., Puigagut, J., 2011. The effect of primarytreatment and flow regime on clogging development in horizontal subsurfaceflow constructed wetlands: an experimental evaluation. Water Res. 45 (12),3579e3589.
Platzer, C., Mauch, K., 1997. Soil clogging in vertical flow reed beds - mechanisms,parameters, consequences and solutions? Water Sci. Technol. 35 (5), 175e181.
Rousseau, D.P.L., 2005. Performance of Constructed Treatment Wetlands: Model-based Evaluation and Impact of Operation and Maintenance. PhD. thesis.Ghent University, Ghent, Belgium, 300 pp.
G.F. Hua et al. / Journal of Environmental Management 119 (2013) 67e75 75
Stasinakisa, A.S., Mamais, D., Thomaidisa, N.S., Lekkas, T.D., 2002. Effect of chro-mium (VI) on bacterial kinetics of heterotrophic biomass of activated sludge.Water Res. 36 (13), 3341e3349.
Tanner, C.C., 1994. Growth and nutrition of Schoenoplectus validus in agriculturalwastewaters. Aquat. Bot. 47 (2), 131e153.
Tanner, C.C., Sukias, J.P., 1995. Accumulation of organic solids in gravel bed con-structed wetlands. Water Sci. Technol. 32 (3), 229e239.
Tanner, C.C., Sukias, J.P.S., Upsdell, M.P., 1998. Organic matter accumulation duringmaturation of gravel-bed constructed wetlands treating farm dairy wastewa-ters. Water Res. 36 (10), 3046e3054.
Winter, K.J., Goetz, D., 2003. The impact of sewage composition on the soil cloggingphenomena of vertical flow constructed wetlands. Water Sci. Technol. 48 (5),9e14.
Wu, Y.G., Hui, L., Wang, H., Li, Y.F., Zeng, R., 2007. Effectiveness of riverbank filtrationfor removal of nitrogen from heavily polluted rivers: a case study of Kuihe River,Xuzhou, Jiangsu, China. Environ. Geol. 52, 19e25.
Wynn, T.M., Liehr, S.K., 2001. Development of a constructed subsurface-flow wet-land simulation model. Ecol. Eng. 16 (4), 519e536.
Yuan, D.H., Ren, Q.J., Gao, S.X., Zhang, H., Yin, D.Q., Wang, L.S., 2004. Purificationefficiency of several wetland macrophytes on COD and nitrogen removal fromdomestic sewage. Appl. Ecol. 15 (12), 2337e2341.
Zhao, Y.Q., Sun, G., Allen, S.J., 2004. Anti-sized reed bed system for animal waste-water treatment: a comparative study. Water Res. 38 (12), 2907e2917.
Zhao, L.F., Zhu, W., Tong, W., 2009. Clogging processes caused by biofilm growth andorganic particle accumulation in lab-scale vertical flow constructed wetlands.J. Environ. Sci. 21 (6), 750e757.
