www.elsevier.com/locate/procbio

Process Biochemistry 43 (2008) 297–305

Domestic wastewater treatment using batch-fed constructed wetland and

predictive model development for NH3-N removal

S.Y. Chan a,*, Y.F. Tsang a, L.H. Cui b, H. Chua a

a Civil & Structural Engineering Department, The Hong Kong Polytechnic University, Hong Kong SAR, Chinab Department of Environment and Resources, The South China Agricultural University, Guangzhou, China

Received 3 September 2007; received in revised form 28 November 2007; accepted 14 December 2007

Abstract

In this study, the performance of a pilot-scale batch-fed constructed wetland in treating domestic wastewater from small community was tested.

The principal of the system capitalizes on the pollutant removal mechanisms of the soil–plant–microbial interactions of constructed wetlands, and

the system operation was integrated with the rhythmical movement of wastewater and air that similar to the operation of conventional sequencing

batch reactor. Based on the hydraulic loading of 0.91 m3/m2/day and the daily maximum contact time of 18 h, the system could achieve around

60% removal efficiency for carbonaceous matters. The removals of ammonia nitrogen and phosphorus were about 50 and 40%, respectively, while

the removal of total suspended solids was approaching 80%. Mathematical models were developed to describe ammonia nitrogen degradation in

the batch-fed constructed wetland. Three analytical approaches including multivariate regression, first-order kinetics and mass balance analysis

were done. Prediction model was formulated to predict the system removal efficiency of ammonia nitrogen.

# 2007 Elsevier Ltd. All rights reserved.

Keywords: Domestic wastewater; Batch-fed; Constructed wetland; Ammonia nitrogen; Predictive model; Coal slag; Cyperus alternifolius

1. Introduction

Domestic wastewater is mainly composed of organic

matters, nutrients and suspended solid. In the treatment process

of domestic wastewater, the removals of organic matters and

nutrients are critical to judge the performance of the treatment

process. The type of wastewater treatment systems being

utilized is a matter of consideration based on the targeted

pollutants to be eliminated, the volume of domestic wastewater

and hence the size of the served population, the local financial

budget and the geographical characteristics [1]. Most waste-

water treatment systems in rural areas with low-dense

population are characterized by low capital investment and

operating cost. Common domestic wastewater treatment

methods in rural areas capitalize on microbial degradation.

They include waste stabilization pond, wastewater storage and

treatment reservoir, upflow anaerobic sludge blanket reactor,

biofilter, aerated lagoon, oxidation ditch and constructed

wetland [2]. However, these treatment systems are limited to

* Corresponding author. Tel.: +852 27666027; fax: +852 23346389.

E-mail address: jackiesychanm@yahoo.com.hk (S.Y. Chan).

1359-5113/$ – see front matter # 2007 Elsevier Ltd. All rights reserved.

doi:10.1016/j.procbio.2007.12.009

application in rural areas with small wastewater flows due to

their large land requirements and relatively less promising

pollutant treatment efficiencies.

In this study, the principle of the treatment system

capitalized on the pollutant removal mechanisms of the soil–

plant–microbial interactions of constructed wetlands. Simulat-

ing a wetland ecosystem, the treatment system can make use of

the assimilation capacity of soil and aquatic plants to remove

both pollutants and nutrients without additional energy demand

[3]. To have prolonged and satisfying pollutant removal

efficiencies, no matter constructed wetlands or other biological

filters need to be rested periodically to allow breakdown of

accumulated organic matter. It is because resting of beds allows

aeration and reduces the likelihood of anoxia. Drying of beds is

occasionally required to enhance performance [4,5]. So the

process design of the studied system has integrated the resting

of bed in a single bed as a time sequence. Different to the

conventional continuous flow bed, the system is a batch-fed

constructed wetland integrated with the rhythmical movement

of wastewater and air like that of a sequencing batch reactor [6].

The operating conditions, such as contact time, temperature and

loading of wastewater to the system, were studied specifically

for the removal of carbon, nitrogen and phosphorus that are

Table 1

Durations of different operating stages

Phase Duration (h)

Fill 4 4 4 4 4

React 18 12 6 3 0

Draw 2 2 2 2 2

Idle 0 6 12 15 18

Total: 24

S.Y. Chan et al. / Process Biochemistry 43 (2008) 297–305298

abundant in domestic wastewater. Predictive models using

multivariate regression, first-order kinetics and mass balance

approaches were also developed for ammonia nitrogen

removal.

2. Materials and methods

2.1. System description

The pilot-scale experiment was carried out onsite in the South China

Agricultural University (SCAU), Guangzhou, People’s Republic of China. A

portion of domestic wastewater from local sewer serving 800 households in the

campus was drained to the batch-fed constructed wetland as influent during the

operation period. Three identical tanks were constructed in parallel near a pond

in the campus of SCAU. Each system was made of concrete with the dimensions

of 5.0 m in length, 3.0 m in width and 1.8 m in depth. Means of experimental

data were calculated from the data from these three identical tanks. Coal slag,

which is the waste residue from burning coal for electricity generation in

Guangdong areas [7], was filled up in the system as the supporting medium. The

uses of this industrial waste or by-product in wastewater treatment served the

purposes of operating economically viable industrial and wastewater treatment

processes for protection of the environment and public health. The porosity and

the density of coal slag were 0.5 � 0.17 and 1.818 � 0.425 kg/L, respectively.

50–100 mm stones were placed around the influent distributor and the effluent

collector pipes to reduce the potential of clogging. The empty bed volume of

each system is 27 m3 with an effective volume of 13.6 m3. The wastewater was

drained into the inlet zone, and then passed through the perforated partition into

the bed matrix. The outlet valves of the systems were at the bottom of the

systems on the other side (Fig. 1). Plants were introduced into the coal slag bed

to imitate the main components of constructed wetland. Clumps of whole plants

of Cyperus alternifolius were planted into the systems with a density of 3–

4 plant/m2. This wetland plant was selected due to its long root system which

can be up to 2 m. The extensive root system could enhance dissolved oxygen in

root zone of planted system and facilitate pollutant degradation process.

2.2. System operations

Different from conventional constructed wetland, the system studied was

a batch reactor. The operating cycle was similar to conventional sequencing

batch reactor, which involves ‘‘fill’’ phase, ‘‘react’’ phase, ‘‘draw’’ phase and

‘‘idle’’ phase [8], but without ‘‘aerate’’ phase. During the operation, waste-

water was pumped into coal slag bed in the ‘‘fill’’ phase and retained in the bed

in the ‘‘react’’ phase. Pollutant removal processes took place in the bed matrix

of such attached-growth system, main removal mechanisms including phy-

sical adsorption and biochemical degradations occurred on the coal slag

surface and the biofilm established on it, as well as certain level of plant

uptake. The batch systems were operated with different durations of ‘‘react’’

Fig. 1. Side view of the batch-fed constructed wetland.

phase (0, 3, 6, 12 and 18 h). Table 1 shows the durations of different phases

accordingly. The systems were operated with a cycle of 24 h from May 2005

to March 2006. The resulting hydraulic loading was at a rate of 0.91 m3/m2/

day. Seasonal variation was considered as the pilot-scale experiment was

carried out in outdoor. In this experiment, the months from May 05 to

September 05 having mean temperature from 22.7 to 25.3 8C were classified

as warm period while from November 05 to March 06 having mean tem-

perature from 9.8 to 19.1 8C were classified as cool period. The five contact

times (‘‘react’’ phase) were tested for their effects on pollutant removal in

both warm and cool periods. Every contact time was tested for a 2-month

period with 1 month in warm period and 1 month in cool period.

2.3. Analytical methods

Samplings of influent and effluent of the systems were done on every run.

Mean of data from three identical systems was calculated for each run.

Analyses of COD, BOD5, NH3-N, TP and TSS were performed following

the Standard Methods [9]. Removal efficiencies were obtained by calculating

the percentages of pollutant removal from the influent concentrations. Tem-

peratures of wastewater were recorded onsite. Dissolved oxygen (D.O.) levels

at different depths of the pilot systems were measured using a portable D.O.

meter (YSI Model no. 51). Statistical and mathematical software including

SPSS 10.0, Datafit 8.0 Okadale and Igor Pro 5.03 were used for model

development and performance predictions. The surface characteristic of coal

slag was observed under Philips XL 30 Esem-FEG Environmental Scanning

Electron Microscope (SEM).

3. Results and discussions

3.1. Overall performance

Table 2 shows the influent characteristics of the raw sewage

entering the systems. According to the classification of Metcalf

and Eddy [10], with the mean concentration of ammonia

nitrogen of 32.71 � 8.37 mg/L, the wastewater was regarded as

having a medium strength of nitrogen content. Before system

operation, approximately 4–5 months were spent in the start-up

period to allow extensive plant root and biofilm development on

Table 2

Means of Influent and effluent characteristics over the operation period

Parameter Mean Standard deviation

pH 7.21 0.27

Sewage temperature (8C) 20.47 4.16

BOD5 (mg/L) 27.59 12.98

COD (mg/L) 80.82 31.52

NH3-N (mg/L) 32.71 8.37

TKN (mg/L) 35.56 9.69

Nitrate (mg/L) 4.85 4.56

TP (mg/L) 2.61 0.99

TSS (mg/L) 23.26 8.73

D.O. (mg/L) 1.40 1.04

Fig. 2. Biofilm establishment on coal slag surface.

Fig. 4. Effects of contact time on different pollutant removal efficiencies in the

pilot-scale system.

Table 3

p-Values of different pollutant removal efficiencies in warm and cool periods

Contact time (h)

0 3 6 12 18

NH3-N 0.023 0.040 0.048 0.000 0.209

BOD5 0.917 0.317 0.034 0.005 0.994

COD 0.290 0.337 0.710 0.219 0.737

TP 0.688 0.372 0.137 0.771 0.248

TSS 0.205 0.458 0.067 0.775 0.758

S.Y. Chan et al. / Process Biochemistry 43 (2008) 297–305 299

the coal slag in the systems. Wastewater was fed into the

system intermittently during the start-up period. The establish-

ment of biofilm on the surface of the coal slag was confirmed by

SEM (Fig. 2) and it reflected the accomplishment of the start-

up period after around 4 months [11]. Steady ammonia

nitrogen concentration in the effluent was also achieved after

around 4 months of start-up period (Fig. 3), the constant

removal showed the existence of regenerative biodegradation

function of the biofilm on the coal slag surface. Based on the

hydraulic loading of 0.91 m3/m2/day and the daily maximum

contact time of 18 h, the system could achieve around 60%

removal efficiency for BOD5. The removals of ammonia

nitrogen and phosphorus were about 50 and 40%, respectively,

while the removal of total suspended solids was approaching

80% (Fig. 4). C. alternifolius has showed satisfactory growth

throughout the study period. The plants have been harvested

once during the middle of the study period. Symbiotic

relationship between the plants and the microorganisms in the

rhizosphere was existed to enhance pollutant degradation [12],

but in-depth investigation of it was out of the scope of this

Fig. 3. Trend of ammonia nitrogen concentration in influent and effluent during

the start-up period.

study. No clogging of substrate was observed around 1-year

operation due to the relatively low concentration of TSS in

the influent that has passed the sedimentation tank as the

pre-treatment.

Removals of COD, BOD5, TP and TSS were not

significantly different in warm and cool periods (ANOVA, p-

values >0.05, n = 100), but the removal efficiencies of NH3-N

were significantly different between warm and cool periods

regardless of the contact times (ANOVA, p-values <0.05,

n = 100) (Bold values in Table 3). Temperature has determinant

effect on the degradation of ammonia nitrogen in the batch-fed

constructed wetland.

3.2. Effects of sequencing batch mode

The availability of oxygen in the filter bed matrix is often

assumed to be the key factor restricting the removal rates of

BOD5, COD and NH3-N [13]. Different pollutant removal

processes are favoured in specific ranges of dissolved oxygen

concentrations. Carbonaceous BOD removal is favoured with

D.O. level of 1–2 mg/L, nitrification is facilitated with D.O.

level of 2–3 mg/L while denitrification is facilitated with D.O.

level <0.5 mg/L. And nitrification took place in all locations

where D.O. levels were higher than the critical threshold of

0.5 mg/L [14]. The average D.O. concentration in the batch-fed

constructed wetland was 2.10 � 1.18 mg/L. With the D.O.

usually over 0.5 mg/L, nitrification was favoured to convert

ammonia nitrogen into nitrate in the batch-fed biofilm reactor

by microbial actions.

S.Y. Chan et al. / Process Biochemistry 43 (2008) 297–305300

The batch-fed constructed wetland was a gravel-based

system integrated with the periodic feeding like that of a

sequencing batch reactor. The air drawn into sequencing batch

beds during the ‘‘draw’’ phase was used as an oxygen source to

biodegrade the pollutants. The oxygen transport and consump-

tion rate in the beds could be greatly improved by the

rhythmical water and air movement in the bed matrix [15].

Moreover, during the ‘‘idle’’ phase, resting of beds allowed air

to get into the bed for aeration and reduce the likelihood of

anoxia. Drying of beds is occasionally required to enhance

performance. These alternating phases of ‘‘feed’’ and ‘‘rest’’

are fundamental to control the growth of the attached biomass

on the adsorbing medium, to maintain aerobic conditions

within the filter bed and to mineralize the organic deposits.

Besides the removal of nitrogen, Miller and Wolf [16] have

shown that P adsorption capacity of vertical filtration bed can be

regenerated if the system is allowed to rest and dry. In other

words, intermittent loadings are feasible to enhance N and P

removals.

3.3. Model development of ammonia nitrogen removal

Fig. 5 shows the ammonia nitrogen removal efficiency with

different contact times, it reflected the extent of nitrification

process achieved in the batch-fed constructed wetland with

regard to the duration of contact time. Ammonia nitrogen

removals were also significantly different in warm and cool

periods of the study (Table 3). In order to find out the optimal

operating conditions of the current system design, mathema-

tical models were developed to predict the precise ammonia

nitrogen removal efficiency with various contact time under

various temperature ranges. Prediction models were developed

based on ammonia nitrogen removal only because ammonia

nitrogen is one of the most dominant pollutants in domestic

wastewater. No equation for phosphorus mass balance was done

because sources and sinks of phosphorus is mainly dependent

on adsorption and desorption. Three types of modelling

including multivariate regression, first-order kinetics and mass

balance were developed, so as to understand the underlying

removal mechanism of ammonia nitrogen through different

Fig. 5. Ammonia nitrogen removal efficiency with different contact times in

warm and cool periods.

approach. Table 4 lists the state variables and their descriptions

in the models.

3.3.1. Multivariate regression model

Since the performance of the batch-fed constructed wetland

was influenced by many factors, such as climate, hydraulic

condition, vegetation, water quality, oxygen level, microbiol-

ogy and influent concentration. Multivariate regression is

attempted for the purpose of finding the latent factors that may

affect and control the system operation and the relationships

between the removals of pollutants and the controlling factors

can be established. While the elimination of ammonia nitrogen

is mainly dependent on nitrification, which is governed by

growth of chemoautotrophic nitrifying bacteria, pH, tempera-

ture, concentration of ammonia nitrogen in the influent and

dissolved oxygen level. BOD5 is also a governing factor of

nitrification because the availability of oxygen within the

system is related to BOD5 concentration, the heterotrophic

bacteria will outgrow the nitrifiers when BOD5 is readily

available [17]. Thus, the elimination of ammonia nitrogen in

the system is then the function of several factors as follows:

ðNH3-NÞout¼ f ððNH3�NÞin;CT;D:O:;Temp;BODinÞ (1)

The multivariate regression model generated by SPSS for

predicting NH3-N concentration in effluent is shown as follows:

ðNH3-NÞout¼ 2:233 þ 0:623ðNH3-NÞin� 0:484CT

þ 0:228D:O:Correlationcoefficient ¼ 0:74 (2)

The model showed that (NH3-N)out was proportional to

(NH3-N)in and dissolved oxygen, but negatively proportional to

contact time. Temperature and BOD concentration did not

significantly affect the nitrification process from the result of

regression analysis of the pilot-scale data. Fig. 6 shows

the fitness of the measured effluent concentrations with the

predicted effluent concentrations. The model predicted

the variability of the ammonia nitrogen in the effluent with

the correlation of 0.74.

The multivariate regression model was used for modeling

the relationship between the controlling factors using a linear

equation. Nonlinear analytical technique—first-order kinetics

was used to develop the next model.

3.3.2. First-order kinetics model

The growth of the nitrifying bacteria is expressed as a first-

order reaction of maximum growth rate and concentration of

organisms [18]. Kadlec [19] described that the kinetics purports

to represent the wetland output concentrations in response to

influent concentration. The first-order rate model, which is a

non-linear relationship, is widely used to design constructed

wetlands and to predict removal performance for pollutants

[20]. The first-order kinetics of ammonia nitrogen removal is as

follows:

dðNH3-NÞdt

¼ �kðNH3-NÞ (3)

Table 4

State variables in the model

State variable Description Units

V Reactor volume L

Q Flow rate L/h

(NH3-N)out Ammonia nitrogen concentration of effluent mg/L

(NH3-N)in Ammonia nitrogen concentration of influent mg/L

CT Contact time h

m Growth rate h�1

mmax Maximum specific growth rate of nitrifier h�1

k Reaction rate coefficient h�1

rA Rate of substrate utilization mg/L h

Ks Half saturation constant of nitrification mg/L

mmax(p) Maximum specific growth rate of plant h�1

Km Half saturation constant of plant uptake of ammonia nitrogen mg/L

Yx/s Yield coefficient of nitrifier mg biomass/mg substrate

Yx/s(p) Yield coefficient of plant mg biomass/mg substrate

Ko Half saturation constant of D.O. mg/L

T Temperature 8Cu Temperature coefficient Dimensionless

D.O. Dissolved oxygen level mg/L

X Biomass concentration mg/L

Xp Plant biomass concentration mg/L

S.Y. Chan et al. / Process Biochemistry 43 (2008) 297–305 301

The first-order kinetics model generated by fitting the ammonia

nitrogen removal data in Datafit 8.0 Okadale is as follows:

ðNH3-NÞout ¼ 0:6867ðNH3-NÞine�0:0243ðCTÞ

Correlation coefficient ¼ 0:79 (4)

The model showed that (NH3-N)out was proportional to (NH3-

N)in, and exponential to contact time. Fig. 7 shows the fitness of

the first-order kinetics model. The R2 value between the

measured and the predicted ammonia nitrogen effluent con-

centrations was 0.75. It showed the certainty that the correlation

was not due to randomness of the data. Controlling factors

Fig. 6. Fitness of predictions from the multivariate regression model.

including temperature, D.O. and BOD5 were excluded by the

modeling process.

3.3.3. Mass balance model

Both multivariate regression and first-order kinetics models

utilized a ‘‘black box’’ approach to predict the ammonia

nitrogen removal in the system. Both models focused on the

overall performance of a system and the major removal

mechanisms were not taken into account. The third model was

developed in this study using mass balance of increased

complexity. A mass balance (also called a material balance) is

an accounting of materials entering and leaving a system [21].

Fig. 7. Fitness of predictions from the first-order kinetics model.

S.Y. Chan et al. / Process Biochemistry 43 (2008) 297–305302

A mass balance model was developed in order to simulate and

investigate the nitrification process occurred in the batch-fed

constructed wetland. Monod’s model which relates the growth

rate of microorganism (nitrifier) to the concentration of a single

growth-controlling substrate (NH3-N)in via two parameters, the

maximum growth rate (mmax) and the substrate affinity constant

(Ks) [13], was adopted here in Eq. (5),

m ¼ mmaxðNH3-NÞinKs þ ðNH3-NÞin

(5)

While the mass flow of ammonia nitrogen in the batch-fed

constructed wetland can be written as in Eq. (6),

(6)

In a batch-growth culture system, a portion of substrate is

converted to new cells. And the growth rate of nitrifier is related

to the rate of NH3-N utilization by nitrifier with a yield

coefficient (Yx/s). Yx/s is defined as the ratio of mass of cell

formed to the mass of substrate consumed, and it is measured

during any finite period of logarithmic growth [10]. In addition,

similar relationship exists with the yield coefficient of plant (Yx/

s(p)) applied to the plant uptake kinetics of ammonia nitrogen in

the batch-fed constructed wetland according to Michaelis–

Menten equation [22,23]. Hence, the overall rate of NH3-N

utilization (rA) in the batch-fed constructed wetland by nitrifiers

and plant uptakes were comprised of 2 components as follows:

Rate of NH3-N utilization by nitrifiers

¼ � 1

Yx=s

mmaxðNH3-NÞinKs þ ðNH3-NÞin

ðxÞ ðMonod0s equationÞ (7)

Rate of NH3-N utilization by plants

¼ � 1

Yx=sðpÞ

�mmaxðpÞðNH3-NÞinKm þ ðNH3-NÞin

ðxpÞ ðMichaelis�Menten equationÞ

(8)

Additionally, seasonal variation of temperature was proved

to have significant effect on nitrification process. The Van’t

Hoff–Arrhenius equation provides a generalized estimate of

temperature effects on biological reaction rates [10]. The

growth rate of Nitrosomonas bacteria also responses to

substrate concentration and dissolved oxygen concentration

according to Monod-type function.

Considering the effects of temperature and dissolved

oxygen, and putting Eqs. (7) and (8) into Eq. (6), the mass

balance model based on the nitrification by nitrifier and the

plant uptake for the removal of ammonia nitrogen in

wastewater was developed as in Eq. (9),

VdðNH3-NÞ

dt¼ QðNH3-NÞin � QðNH3-NÞout

� V

�1

Yx=s

mmaxðNH3-NÞinKs þ ðNH3-NÞin

ðxÞ��

D:O:

Ko þ D:O:

�uT�20

ðNitrificationÞ

� V

�1

Yx=sðpÞ

mmaxðpÞðNH3-NÞinKm þ ðNH3-NÞin

�ðxpÞ

ðPlant uptakeÞ

(9)

And for batch reactor,

Q ¼ 0 and V is constant; (10)

And

dðNH3-NÞdt

¼ DðNH3-NÞDt

¼ ðNH3-NÞout � ðNH3-NÞinCT

(11)

Putting Eqs. (10) and (11) into Eq. (9), the model was modified

as follows in Eq. (12):

ðNH3-NÞout ¼ ðNH3-NÞin � CT

�1

Yx=s

mmaxðNH3-NÞinKs þ ðNH3-NÞin

ðxÞ�

��

D:O:

Ko þ D:O:

�uT�20

��

1

Yx=sðpÞ

mmaxðpÞðNH3-NÞinKm þ ðNH3-NÞin

�ðxpÞ ¼ CT

(12)

The final valid model obtained by fitting experimental data into

Eq. (12) with Datafit 8.0 Okadale is as follows:

ðNH3-NÞout ¼ ðNH3-NÞin � ðCTÞ�

1

Yx=s

mmaxðNH3-NÞinKs þ ðNH3-NÞin

ðxÞ�

��

D:O:

Ko þ D:O:

�uðT�20Þ

Correlation coefficient ¼ 0:92 (13)

The final mass balance model (Eq. (13)) indicated that the

most important factors affecting growth rate of nitrifier were

substrate concentration in the influent, contact time, dissolved

oxygen and temperature. The effect of plant uptake was

excluded in the final mass balance model. The mass balance

model development found that the plant uptake function

represented by Michaelis–Menten equation was not significant

in the ammonia nitrogen degradation process in the batch-fed

constructed wetland. Even though, it may not be appropriate to

totally neglect the role of plant in the bed system, as apart from

Table 5

Coefficients of nitrification in the mass balance model

Coefficients mmax (day�1) Ks (mg/L) Ko (mg/L) u

Value in the model 0.0064 8.947 0.026 1.032

Typical value 0.008 1.3 0.2 1.04

Metcalf and Eddy (2003).

Table 6

Results of sensitivity analysis

Parameters Units Assigned range Assigned value Perturbation Sx

(NH3-N)in mg/L 14.45–51 32.73 +1% 1.323(+)

�1% 1.322(+)

Average 1.323(+)

CT h 0–18 9 +1% 0.336(�)

�1% 0.336(�)

Average 0.336(�)

T 8C 14.47–28.61 21.54 +1% 0.280(�)

�1% 0.278(�)

Average 0.279(�)

D.O. mg/L 0.2–3.28 1.74 +1% 0.006(�)

�1% 0.006(�)

Average 0.006(�)

(+) indicates a positive relation between the changes in the parameters and the

change in the model output. (�) indicates an inverse relation between the

changes in the parameters and the change in the model output.

S.Y. Chan et al. / Process Biochemistry 43 (2008) 297–305 303

direct plant uptake, plants could facilitate the pollutant removal

processes by possible oxygen release from the root zone to the

surrounding adsorbing medium [24]. Also, additional surface

area for microbial attachment is also provided by plant in

wetland systems [25]. In some plants, degradation of

contaminants occurs when root exudates (e.g. simple sugars,

alcohols and acids) stimulate proliferation of microbial

communities in the rhizosphere. This is known as rhizo-

enhanced degradation. Roots also de-aggregate the soil matrix,

allowing aeration and promoting biodegradation [26]. Plant is

still regard as an important component in such filter bed system.

The correlation coefficient of the mass balance model was

0.92, which was the highest among the three types of models

being tested. Mass balance approach is able to determine the

sources and sinks of nitrogen in the system, and allows the

understandings of the types of removal mechanisms involved as

well as their significance. Referring to the final mass balance

model in Eq. (13) and in order to increase the removal of

ammonia nitrogen in the batch-fed constructed wetland, factors

including influent concentration, contact time, dissolved

oxygen and temperature have to be considered in priority.

Table 5 shows the typical kinetics coefficients of the

nitrification process with the comparison of the values obtained

from the mass balance model. The biomass (x) of a typical fixed

bed reactor from literature is 800 mg/L [18] while the typical

value of Yx/s in nitrification process is 0.2 [10]. The calculated

Fig. 8. Fitness of the predicted values from the mass balance model.

value of mmax in the batch-fed constructed wetland was 0.0064

day�1. The value was comparable to the typical values in

literature. Fig. 8 shows the fitness of the measured effluent

concentration with the predicted effluent concentration from

the mass balance model. The R2 value of the fitness was 0.75.

The findings were as good as those in other similar modelling

studies using data from pilot-scale experiment with the range of

R2 value from 0.7 to 0.8 [27].

3.3.4. Sensitivity analysis

The sensitivity of the ammonia nitrogen mass balance model

was tested using the four parameters: influent ammonia

nitrogen concentration (NH3-N)in, contact time (CT), tempera-

ture (T) and dissolved oxygen (D.O.) (Table 6). Except the

concentration of ammonia nitrogen in influent showing a

positive relationship with the model output, contact time,

temperature and dissolved oxygen were all showing a negative

relationship with the model output. An overview of the most

sensitive components in the model can be obtained through the

sensitivity analysis [28]. The sensitivity of the influent

ammonia nitrogen concentration was the largest at +1.323

(+1%) and +1.322 (�1%), so the model was the most sensitive

to it. The Sx value of contact time was bigger than that of

Fig. 9. Contour plot for ammonia nitrogen removal prediction at sewage

temperature in a range of 9.8 8C < T < 20 8C.

Fig. 10. Contour plot for ammonia nitrogen removal prediction at sewage

temperature in a range of 20.0 8C � T < 35 8C.

S.Y. Chan et al. / Process Biochemistry 43 (2008) 297–305304

temperature, but the model was less sensitive to these two

parameters than to the influent ammonia nitrogen concentra-

tion. Lastly, the model was the least sensitive to dissolve

oxygen as the Sx value was only about 0.006 (�1%).

3.4. Applications

To cope with different wastewater demands and desired

effluent qualities, contour plots were prepared from the mass

balance predictive model. The predictions of treatment

efficiencies were done for two temperature ranges, with regard

to the recorded temperature range in the studied areas. Figs. 9

and 10 show the predicted removal efficiencies of ammonia

nitrogen at sewage temperature in a range of

9.8 8C < T < 20 8C and in a range of 20.0 8C � T < 35 8C,

respectively. The required contact time for achieving desired

levels of removal efficiencies could be predicted under different

influent concentrations.

4. Conclusions

The batch-fed constructed wetland was a modified biofilter

system that integrated with sequencing batch-feeding mode and

with the presence of plants. The system could achieve desirable

removal efficiencies of BOD5, NH3-N, TP and TSS in the pilot-

scale experiment of this study. The predictive models

developed in this study provide guidance to system design

for different wastewater treatment purposes and demands with

different ammonia nitrogen loadings and desired removal

efficiencies in suburban areas. With its ease of operation and

low-cost requirement, the system provides an alternative

domestic wastewater treatment tool in suburban populated

area where is out of the coverage of the municipal sewage

treatment network. Moreover, the selection of plant species and

supporting media in the system can be varied according to

different local characteristics, in order to fulfil both environ-

mental and financial considerations in different areas.

Acknowledgments

The Hong Kong Polytechnic University Research Grant and

the Hong Kong Research Grants Council are hereby acknowl-

edged for the financial supports.

References

[1] Mitsch WJ, Gosselink JG. Wetlands. New York: Van Nostrand Reinhold;

1993.

[2] Mara D. Domestic wastewater treatment in developing countries. London:

Earthscan; 2003.

[3] Inamori R, Gui P, Dass P, Matsumura M, Xu KQ, Kondo T, et al.

Investigating CH4 and N2O emissions from eco-engineering wastewater

treatment processes using constructed wetland microcosms. Process

Biochem 2007;42:363–73.

[4] Okurut TO, Rijs BGJ, Bruggen JJA. Design and performance of experi-

mental constructed wetlands in Uganda, planted with Cyperus papyrus

and Phragmites mauritianus. Water Sci Technol 1999;40(3):265–71.

[5] Platzer C. Design recommendations for subsurface flow constructed

wetlands for nitrification and denitrification. Water Sci Technol

1998;40(3):257–63.

[6] Morgenroth E, Wilderer PA. Sequencing batch reactor technology—

concepts, design, experience. J CIWEM 1998;12(5):314–21.

[7] Cui LH, Zhu XZ, Luo SM, Liu YH. Purification efficiency of

vertical-flow wetland system constructed by coal slag and turf adsorbing

medium on municipal wastewater. Chin J Appl Ecol 2003;14(4):

597–600.

[8] Obaja D, Mace S, Costa J, Sans C, Mata-Alvarez J. Nitrification, deni-

trification and biological phosphorus removal in piggery wastewater using

a sequencing batch reactor. Biores Technol 2003;87:103–11.

[9] APHA. Standard methods for the examination of water and wastewater,

20th ed., Washington, DC: American Public Health Association; 1998.

[10] Metcalf and Eddy, Inc.. Wastewater engineering: treatment disposal and

reuse, 4th ed., New York: McGraw-Hill Inc.; 2003.

[11] Paul EA, Clark FE. Soil microbiology and biochemistry. USA: Academic

Press; 1988.

[12] Tanner CC. Plants as ecosystem engineers in subsurface-flow treatment

wetlands. Water Sci Technol 2001;44(11–12):9–17.

[13] Sun G, Gray KR, Biddlestone AJ, Allen SJ, Cooper DJ. Effect of effluent

recirculation on the performance of a reed bed system treating agricultural

wastewater. Process Biochem 2003;39:351–7.

[14] Taylor T, Bishop P. Distribution and role of bacterial nitrifying population

in nitrogen removal in aquatic treatment system. Water Res 1989;23:

947–55.

[15] Sun G, Gray KR, Biddlestone AJ, Cooper DJ. Treatment of agricultural

wastewater in a combined tidal flow–downflow reed bed system. Water Sci

Technol 1999;40(3):139–46.

[16] Miller FP, Wolf DC. Renovation of sewage effluents by soil. In: Individual

onsite systems. Second National Conference, National Sanitation Foun-

dation; 1975.p. 87–102.

[17] Horan NJ. Biological wastewater treatment systems—theory and opera-

tion. England: John Wiley & Sons; 1990.

[18] Halling-Sørensen B, Nielsen SN. A model of nitrogen removal from

wastewater in a fixed bed reactor using simultaneous nitrification and

denitrification (SND). Ecol Model 1996;87:131–41.

[19] Kadlec RH. Wetlands for wastewater polishing: free water surface

wetlands. In: Mitsch WJ, editor. Global wetlands: old world and new.

Amsterdam: Elsevier; 1994. p. 335–49.

[20] Kim SJ, Hong SW, Choi YS, Bae WK, Lee SH. Performance evaluation of

a newly developed flow diverted bed system for stream restoration.

Process Biochem 2007;42:199–209.

[21] Crites R, Tchobanoglous G. Small and decentralized wastetwater manage-

ment systems. In: McGraw-Hill series in water resources and environ-

mental engineering. USA: WCB McGraw-Hill; 1998.

[22] Marcus-Wyner L, Rains DW. Simultaneous measurement of NH4+ absorp-

tion and N2 fixation by Glycine max L., Response to temperature, pH and

external nitrogen concentration. Plant Physiol 1982;69:460–4.

[23] Tylova-Munzarova E, Lorenzen B, Brix H, Votrubova O. The effects of

NH4+ and NO3

� on growth, resource allocation and nitrogen uptake

kinetics of Phragmites australis and Glyceria maxima. Aquat Bot

2005;81:326–42.

[24] Brimblecombe P, Jickells TD, Liss PS, Reid B. An introduction to

environmental chemistry. UK: Blackwell Science Publishing; 2004.

S.Y. Chan et al. / Process Biochemistry 43 (2008) 297–305 305

[25] Gersberg RM, Elkins BV, Goldman CR. Use of artificial wetlands to

remove nitrogen from wastewater. J Water Pollut Control Fed 1984;56:

52–6.

[26] East Texas Plant Materials Center. Constructed wetlands for on-site septic

treatment. A guide to selecting aquatic plants for low-maintenance micro-

wetland. USA: East Texas Plant Materials Center; 2004.

[27] Senzia MA, Mayo AW, Mbwette TSA, Katima JHY, Jørgensen SE.

Modelling nitrogen transformation and removal in primary facultative

ponds. Ecol Model 2002;154:207–15.

[28] Jørgensen SE, Bendoricchio G. Fundamentals of ecological modeling. In:

Developments in environmental modeling3rd ed., Amsterdam: Elsevier;

2001.