Science of the Total Environment 645 (2018) 156–169

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Design and development of a hybrid macrophyte assisted vermifilter for

the treatment of dairy wastewater: A statistical and kinetic
modelling approach
Kundan Samal a,b, Rajesh Roshan Dash a,⁎, Puspendu Bhunia a
a
School of Infrastructure, Indian Institute of Technology Bhubaneswar, 752 050, Odisha, India
b
School of Civil Engineering, Kalinga Institute of Industrial Technology-Deemed to be University Bhubaneswar, 751024, Odisha, India

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Earthworm Eisenia fetida and macro-

phyte Canna indica have been used to de-
velop macrophyte assisted vermifilter
(MAVF).
• Vertical flow (VF) unit promote nitrifica-
tion and horizontal flow (HF) unit pro-
mote denitrification.
• Earthworm and plant maintained requi-
site dissolved oxygen (DO) in filter bed.
• Stover-Kincannon model showed high
regression coefficient value.

a r t i c l e i n f o a b s t r a c t

Article history: Global urbanization, exponential increase in population and sophisticated life style of the present generation are
Received 16 April 2018 the major causes leading to a rapid increase in water demand in recent years. In order to nullify this rising water
Received in revised form 19 June 2018 demand, it's high time to reuse domestic as well as industrial effluent after providing suitable chemical/biological
Accepted 9 July 2018
treatment. Macrophyte filter incorporated with earthworm is identified as one of the most economic system for
Available online xxxx
the treatment purpose in developing countries. However, very few literatures and technical information are
Editor: Paola Verlicchi available to scale up the design and its easy operation. This paper aims to develop a hybrid system and assess
its performance for the purification of dairy wastewater. In the present study, two stage macrophyte assisted
Keywords: vermifilters (MAVFs) have been designed. The 1st stage encompassed a vertical flow (VF) unit, and the 2nd
Vermifilter stage contained a horizontal flow (HF) unit. Both the units were inoculated with earthworm Eisenia fetida and
Macrophyte assisted vermifilter were planted with Canna indica. Box-Behnken model was applied to design the system and study the effect of
Canna indica various parameters. It was observed that hybrid MAVF system removed a maximum of 83.2% COD and 57.3%
Eisenia fetida TN at HLR 0.6 m/d and an active layer depth of 30 cm. Ammonification and nitrification typically occurred in
Vertical and horizontal flow
the active layer (earthworm inoculated zone) of VF unit due to high activity of earthworms and its associated
Kinetic modelling
gut microbes, whereas HF unit facilitate denitrification process. Earthworm growth characteristics in the system
were monitored, which is an imperative factor for the design of MAVF reactor. Kinetic modelling of 1st order, grau
2nd order and Stover-Kincannon model were performed and the Stover-Kincannon model showed high regres-
sion coefficient (COD, R2 0.9961 and TN, R2 0.9353) supporting its applicability as compared to the other models.
© 2018 Published by Elsevier B.V.

⁎ Corresponding author at: School of Infrastructure, Indian Institute of Technology Bhubaneswar, Odisha 752 050, India.
E-mail address: rrdash@iitbbs.ac.in (R.R. Dash).

https://doi.org/10.1016/j.scitotenv.2018.07.118
0048-9697/© 2018 Published by Elsevier B.V.
 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169
1. Introduction
The rapid urbanization in most of the developing countries is mas-
sively contaminating the existing water resources in the recent years.
The industrial growth and various changes in manufacturing processes
increase the generation and complexity of wastewater. Effluents from
various industries may also contain toxic pollutants, which have to be
reduced to protect the environment and public health (Dan et al.,
2011; Vymazal, 2007). Numerous problems are encountered by indus-
trial wastewater discharges such as hydraulic overload, temperature ex-
treme, excessive amounts of FOG (fat, oil and grease), acidic or alkaline
constituent, suspended solid, and inorganic or organic content, etc.
(Demirel et al., 2005; Arias et al., 2005). A vastly organic rich effluents
are produced from food industry, paper plants, wool processing indus-
try, pharmaceutical industry, petrochemicals and refinery, dairy indus-
try, brewery and fermentation factory, dye industry, etc. (Sinha et al.,
2008; Samal et al., 2017a; Borghei et al., 2008; Carballeira et al., 2017;
Mansouri et al., 2014). Wastewater from dairy industry has detrimental
impact on environment due to its high concentration of biodegradable
pollutants. It is calculated that 6–10 L of waste effluents are generated
per litre of milk produced. Carbohydrate, protein and fat content of
wastewater degrade rapidly and get converted into odorous sludge,
which has a lot of ill effects on natural water bodies like increase in tur-
bidity, toxicity, heat retention, pH imbalance, eutrophication, dissolve
oxygen reduction, etc. (Rodgers et al., 2006; Samal et al., 2017b). Several
physical, chemical and biological treatment processes like filtration,
sedimentation, desalination, membrane technology, sonication, ozona-
tion, advanced oxidation process, electrocoagulation, activated sludge,
rotating biological contactors, aerated lagoon, stabilization pond, and
constructed wetland are available. However, challenges still exists due
to environmental, economic, social and legal concerns (Arias et al.,
2005; Borghei et al., 2008; Mansouri et al., 2014; Rodgers et al., 2006).
Vermifilter (VmF) has the potential to treat high organic matters
present in wastewater. It involves the inoculation of earthworms in tra-
ditional geo-filter to further enhance the efficiency of wastewater treat-
ment. Earthworms stimulate and accelerate the microbial activity by
increasing the population of soil microorganisms within filter bed. The
symbiotic activity of earthworms and microorganisms fragment the
pollutants that exist in wastewater. The fragmented particles have
higher surface area, which quicken the biodegradation process. Planting
macrophyte on VmF bed significantly improves the efficiency of bio-
filter, as the microbial community propagates on the root system and
speed up the decomposition process (Singh et al., 2017; Bajsa et al.,
2003). The majority of previous studies are either on utilization of
Fig. 1. Diagrammatic representation of a Macrophyte Assisted Vermifilter (MAVF).
157
vermifiltration or only on macrophyte filtration system. However, not
much comprehensive reports are available on vermifilter integrated
with macrophyte filter or macrophyte assisted vermifilter (MAVF) to
treat wastewater in an effective manner (Fig. 1).
One-Variable-At-a-Time (OVAT) is a conventional methodology for
calculating the effect of various parameters and their interaction. If the
experimental parameters are increased, number of experiments also in-
crease for a particular process, which is both, time consuming and hectic
process. Response surface methodology (RSM) is a convenient tool in
which individual effect of various parameters and their interaction can
be observed simultaneously by performing less number of experiments.
Therefore, the major objective of this study includes designing and de-
velopment of an eco-friendly, low cost and sustainable bio-filter for
treatment of various types organic and nutrient rich wastewater. It
can be achieved through following steps: (i) investigation of the organic
and nitrogen removal efficiency in hybrid MAVF, (ii) application of RSM
to optimize experimental parameters and subsequently develop statis-
tical models for COD and TN removal in hybrid MAVF, (iii) determina-
tion of pollutants degradation kinetics and mechanisms in filter bed;
and (iv) analysis of earthworm and plant growth characteristics.
2. Materials and methodology
2.1. Design of reactors
Two stage laboratory scale reactors were designed made up of
persplex and arranged as shown in Fig. 2 and Fig. 3. Raw wastewater
pumped to the 1st vertical flow (VF) unit at different hydraulic loading
rate (0.3–0.9 m3 m−2 d−1) from a feeding tank and the effluent again
transferred to the 2nd stage horizontal flow (HF) unit. Table 1 displays
the major assemblies of the system. Garden soil and vermicompost
mixed in the ratio of 1:3 by volume and filled in the upper layer of VF.
This layer is called worm active layer as earthworm (Eisenia fetida) are
inoculated in this zone at a density of 10,000 worms/m3 and height of
this layer varied from 15 to 45 cm. Eisenia fetida was selected because
of its tolerance towards highly moistened environment and their higher
reproductive ability than the other species applied in vermi-
remediation. It is even reported that Eisenia fetida can survive water-
logged conditions for certain period of time (Arora et al., 2016; Hait
and Tare, 2011). Being surface-dweller in nature, Eisenia fetida is pre-
ferred over endogeic and anecic earthworm species. The 2nd, 3rd and
4th layers from the top were taken as washed sand (10 cm thick),
4–6 mm laterite stone (15 cm thick) and 8–10 mm coarse gravel
(15 cm thick), respectively. The HF unit bed was completely filled
158 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169

Fig. 2. Design and specification of a hybrid MAVF.

with soil and inoculated with earthworm Eisenia fetida. Both VF and HF
unit are planted with macrophyte Canna indica. It is a perennial plant
producing clumps of stems 150–300 cm tall with large leaves up to
50 cm long and 20 cm wide. The stems arise from a large, thick and
tuber-like rhizome. This plant forms large dense clumps, particularly
in riparian zones (banks of watercourses, wetland edges swamp and
other moist areas), and replaces different wetland species (Samal
et al., 2017a). The detail arrangement of different level of filter bed is ex-
emplified in Table 2. Initial porosity of the media was calculated by
pouring a known amount of water in the bed materials. Wastewater
flowed from top to bottom zone in VF unit, whereas in HF unit flow
was carried on completely in the bottom zone. 1.5% slope was set
from inlet to outlet to facilitate the flow in HF unit. Single-stage reactor
(VF or HF) cannot achieve high removal of total nitrogen due to their in-
ability to provide both aerobic and anaerobic conditions at the same
time. VF unit remove NH+ 4 -N successfully, but very limited denitrifica-
tion takes place in this system. It aids the nitrification process due to
dominance of oxygenated environment. On the other hand, HF unit pro-
vide good conditions for denitrification, but the ability of this system to
nitrify ammonia/organic nitrogen is very limited (Vymazal, 2007).
Therefore, both VF and HF unit combined with each other in order to
exploit the specific advantages of the individual system (nitrification-
denitrification). 25 days of adaptation period were given to all the filtra-
tion system and each experiment was performed for another 80 days.

Fig. 3. Laboratory scale experimental set up.

 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169 159

Table 1 using response surface regression which is given by the following sec-
Specification of different unit of MAVF. ond order polynomial equation:
Specification 1st stage 2nd stage

Flow Vertical subsurface Horizontal subsurface

Shape Circular reactor Rectangular reactor
S ¼ b0 þ b1 x1 þ b2 x2 þ b3 x3 þ b11 x1 2 þ b22 x2 2 þ b33 x3 2 þ b12 x1 x2
Dimension⁎ H 90 cm, D 19.8 cm L 60 cm, W 18 cm, H 30 þ b13 x1 x3 þ b23 x2 x3 ð2Þ
Working H 55–85 cm cm
Working H 25 cm
Plants and Canna indica, Eisenia fetida Canna indica, Eisenia where S is the percentage removal and b0 is the value of fitted response
earthworms fetida
at the center point of design. b1…3, b11…33 and b12, 13, 23 are the linear,
Bed materials Mixture of compost and garden soil, Mixture of compost and
washed sand, laterite, coarse gravel garden soil, laterite quadratic and interaction terms respectively, and x1…3 are dimension-
less coded variables.
⁎ H = Height, D = Diameter, L = Length, W = Width.
A second order polynomial model may be appropriate to establish
the mathematical relationship between the response and the factors
2.2. Composition of wastewater for a 3-factor, 3-level design. The adequacy of the model and the signif-
icant variables affecting the removal efficiency were expressed in terms
Synthetic dairy wastewater was prepared daily one hour before of F-value from Fisher's F-test and probability (p-value) for a 5% signif-
application by addition (per litre of solution) of milk powder icance level. The fitting of the developed model was expressed in
(Amulspray), NaHCO3, NH4Cl, MgSO4.7H2O, KH2PO4, CaCl2·H2O, terms of coefficient of determination (R2), adjusted and predicted R2.
FeCl3.6H2O, MnSO4.7H2O in tap water (Rodgers et al., 2006). Wastewa-
ter containing various concentration of COD (1200–2800 mg L−1) and
2.5. Kinetic modelling
TN (64–148 mg L−1) were poured to the reactor and flow was main-
tained under gravity.
Mathematical models are in general used to optimize the plant de-
sign, predict the operational performance, and to evaluate the experi-
2.3. Analytical procedure
mental results. Kinetic analysis is an accepted pathway for describing
− the performance of biological treatment systems and for predicting
The wastewater samples were analyzed for COD, NH+ 4 -N, NO3 -N,
their performance (Akhbari et al., 2012). Various mathematical models
TN, pH and the methods used to determine above parameters were po-
are available in literature, which have been used to bridge a relationship
tassium dichromate method, nessler reagent method, sodium salicylate
between different process parameters and to scrutinise the experimen-
method, persulfate digestion method, pH electrode, respectively (APHA,
tal data. The performance of MAVF reactors can be analyzed by applying
2005). The efficiency of pollutants was calculated by using following
following selective kinetic models:
formula:

Ci −Co 2.5.1. First-order model

Percentage Reduction ¼  100 ð1Þ
Ci The rate of change in substrate concentration in the system assum-
ing the first order model for substrate removal could be expressed as
where, Ci and C0 symbolize the influent and effluent value, respectively. follows (Isik and Sponza, 2005):

2.4. RSM based experimental design

ds Q Si Q Se
− ¼ − −K 1 Se ð3Þ
The experimental design as a function of the prime influencing fac- dt V V
tors (independent variables) has to be determined in order to find out
the effect of various operating parameters and the optimum conditions
for organics and nitrogen degradation in a MAVF system. The perfor- where, Si = Influent substrate conc. (g L−1)Se = Effluent substrate conc.
mance and efficiency of an optimization model are majorly dependent (g L−1)Q = Influent flow rate (L d−1)V = Reactor volume (L)K1 = First-
on the choice of design of experiment (DoE). The Box-Behnken (BBD) order kinetic constant (d−1).
design provides good accuracy and prediction in the overall experimen- Under pseudo steady state conditions, the rate of change in substrate
tal system. The experimental design consists of 15 experimental runs concentration is negligible ð− ds
dt
Þ ¼ 0, Eq. (3) can be modified as:
(12 experimental and 3 central run) and these experiments enable to
determine the effect of each independent process variables (HLR, pol-
Si −Se
lutants concentration and worm active layer depth) and the interaction ¼ K 1 Se ð4Þ
ƟH
between these variables on the dependent variables (COD, TN). A soft-
ware Design Expert 7.0.0 version (Stat Ease Inc., USA) was employed
to perform analysis of variance (ANOVA) and to generate response sur-
where, ƟH = hydraulic retention time ¼ QV:
faces. The actual values and coded values (in the range of +1 to −1) are
shown in Table 3. The range of actual value was decided based on pre- Thus, the value of K1 can be achieved from the slope by plotting SiQ−S
H
e

liminary experimental results. The experimental data were evaluated versus Se.

Table 2
Filter materials used in MAVF.

Levels Bed materials Role Height (cm) Particle size (mm)

1 Vermicompost: garden soil:: 1: 3 Earthworm habitation 15–45 0–1.18

2 Sand Solids trapped 10 0–2
3 Laterite stone Phosphorous adsorption, supporting layer 15 4–6
4 Gravel Facilitate wastewater flow, biofilm formation zone 15 8–10
160 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169

Table 3 rate (Optaken, 1982; Jin and Zheng, 2009). The original model is repre-
Factors and their levels for BBD. sented by Eq. (11).
Variables Unit Symbol Range and their levels
(coded) ds Q
¼ ðS −Se Þ ð11Þ
−1 0 +1
dt V i

COD where ds is defined as follows:

dt
Hydraulic loading rate (HLR) m/d A 0.3 0.6 0.9
Strength of the wastewater (COD) mg L−1 B 1200 2000 2800

Active layer depth in VF cm C 15 30 45
Q Si
U max
ds V
TN ¼
 ð12Þ
dt Q Si
Hydraulic loading rate (HLR) m/d X 0.3 0.6 0.9 Ks þ
TN concentration mg L−1 Y 64 106 148 V
Active layer depth in VF cm Z 15 30 45
1 V KsV 1

¼ ¼ þ ð13Þ
ds Q ðSi −Se Þ U max QSi U max
dt

2.5.2. Grau second-order model If V

½Q ðSi −Se Þ, which is the inverse of the total substrate removal rate
The general equation of the second-order multicomponent kinetic
plotted against the inverse of the total substrate loading rate ðQS
V
Þ, it
model is exemplified as follows (Borghei and Hosseiny, 2002): i

will give a straight line. The intercept and the slope of this line will

2
ds Se
1
give U max and UKmax
s
, respectively and from those the kinetic constants
− ¼ K2X ð5Þ
dt Si Umax and Ks can be determined.
The substrate balance for the reactor at steady-state can be written
Integrated and linearized form of Eq. (5) is expressed as below ex- as follows:
pression
ds
QSi ¼ Q Se þ V ð14Þ
Si S dt
ƟH ¼ ƟH þ i ð6Þ
Si −Se K2X
Substituting of Eq. (12) into Eq. (14) gives
where, K2 = Second-order kinetic constant (d−1)X = Avg. biomass

conc. in system (g VSS L−1)K2 is the substrate removal rate for each Q Si
U max
unit of microbes. V
QSi ¼ QSe þ V
 ð15Þ
The converted form of Eq. (6) is as below: Q Si
Ks þ
V
V V 1
¼ þ ð7Þ
Q ðSi −Se Þ Q Si K 2 X This expression can be solved for effluent substrate concentration as
follows:
Assuming 2nd term of right hand side of Eq. (6) is constant, the ob-
tained equation is U max Si
Se ¼ Si −
 ð16Þ
QSi
Ks þ
Si V
ƟH ¼ a þ bƟH ð8Þ
Si −Se

Si  −1  3. Results and discussion

where; a ¼ d ð9Þ
K2X
b = constant for Grau second-order model
Also, the substrate removal efficiency, E = Si−SiSe
Hence, the Eq. (6) can be written as:
ƟH
¼ a þ bƟH
E (Liu et al., 2012). There is a symbiotic and synergetic relation exist be-
Kinetic coefficients ‘a’ and ‘b’ can be obtained from the graph
made with θH as x-axis and QEH as y-axis. The intercept and slope of
the plot represents ‘a’ and ‘b’ value, respectively and K2 can be cal-
culated from Eq. (9). The coefficient ‘b’ in Eq. (10) is close to one
and generally reflects the impracticality of attaining a zero value
of COD.
2.5.3. Stover-Kincannon model
The Stover-Kincannon model was initially used to predict the at-
tached growth biomass performance in a rotating biological contactor.
The model was modified and widely applied to describe and predict
the performance of other types of bioreactors. In this model, the
substrate utilization rate is expressed as function of the organic loading
3.1. Organics removal from wastewater
Suspended solids present in the influent are sieved out and depos-
ited on top layer (worm active zone) of the VF unit of the filter and
earthworms present in that layer consume, digest and assimilate
those deposited organic solids with association of soil and gut microbes
ð10Þ
tween earthworms and microbes and Canna indica root system provides
surface area for the growth of pollutants degrading microbes (Samal
et al., 2017b). The digested solids excrete on bed in the form of cast
rich in microbial population. These are more compact, granular and po-
rous aggregates with high surface area and addition of earthworm's
body fluid makes it stickier in nature. Solid particles present in waste-
water adsorb on the surface of worm cast (Bajsa et al., 2003). Overall,
the degradation process occurred in two stages; in first stage, pollutants
are homogenized by earthworms and the generated solid particles are
very tiny size having higher surface area. Those particles are further de-
graded by microbes in second stage and are converted to CO2 and H2O.
Atmospheric oxygen diffuses into the filter bed and roots of Canna indica
releases some amount of oxygen to their rhizosphere which favours the
growth of aerobic microbes. Earthworm burrowing activity also speeds
 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169 161

Fig. 4. Mechanisms of pollutants removal in MAVF reactor.

up atmospheric oxygen diffusion process to the filter bed. Dissolved ox-
ygen present in wastewater and biofilm are also other sources of oxygen
(Samal et al., 2018; Xing et al., 2010). Release of root exudates like cit-
rate, malate, oxalate, acetate, etc. (act as carbon sources) from plants
and earthworm's enzymes accelerate the biodegradation process
(Kantawanichkul et al., 2009). The mechanisms of pollutants removal
in MAVF system are presented in Fig. 4 concisely. Almost half of the
COD removed in VF unit of the reactor was removed in active layer of
the reactor in case of all experimental runs. 37.1% COD was removed
in active layer of VF unit in the case of experimental run 11 and next
to this, 36.8 and 35.5% COD were removed in active layer of
experimental run 13 and 14, respectively. 68.5% COD was treated in
the VF unit of the reactor in experimental run 14 which is highest
among all the runs and in the same run, 83.2% COD was removed at
the final stage (from HF unit). The experimental conditions of run 14
consisted of influent COD 2000 mg L−1, HLR 0.6 m/d and an active
layer depth 30 cm. Major fraction of the organics were removed in VF
unit (45.6% - 68.5% COD) itself in all the experimental run. Active layer
of VF unit plays major role in removing organics as this is the habitation
zone for earthworms and various aerobic soil microorganisms. Fig. 5
shows the residual COD concentration in active layer effluent, VF efflu-
ent and HF (VF + HF) effluent.

Fig. 5. Effluent COD concentration in various experimental run.

162 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169

−
Fig. 6. Effluent (a) NH+
4 -N, (b) DO and (c) NO3 -N concentration in various experimental run.
 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169 163

3.2. Nitrogen removal from wastewater of nitrifying bacteria and ammonium oxidizing bacteria (AOB) was high
and mineralization process was also higher. HF unit facilitate denitrifica-
Major nitrogen elimination processes in MAVF systems are ammoni- tion process due to dominance of anoxic condition. Concentration of
fication, nitrification-denitrification, microbial assimilation, plant up- NO−3 -N in the effluent of HF unit was in between of 8.7 (Run 1) to 26.1
take, matrix adsorption, etc. (Dan et al., 2011). The nitrogen in (Run 10) mg L−1. Combination of nitrification and denitrification process
−
wastewater is primarily present in four forms; Org.-N, NH+ 4 -N, NO2 -N removes TN from wastewater and the end product of denitrification is N2,
and NO− 3 -N. The organic nitrogen present in the wastewater gets con- N2O and NO, which were released to atmosphere.
−
verted to NH+ 4 -N and subsequently gets converted to NO2 -N and There was no significant difference in the effluent pH of MAVF. Ac-
NO− 3 -N by nitrifying bacteria. The nitrified end products further gets cording to previous reports earthworm acts as buffering agent and neu-
denitrified in an anoxic environment and released as atmospheric nitro- tralizes pH in different vermiremediation process. In all experimental
gen (Vymazal, 2007). However in this experiment, NO− 2 -N has been runs, initially the influent pH was 6.34 ± 0.5 and effluent pH of MAVF
found to be nominal in the influent and effluent as well. The reason be- were 7.9 ± 0.3. It may be due to active microbial degradation of depos-
hind the absence of NO− 2 -N could be the easy conversion of the NO2 -N
−
ited organic substances in worm active zone of filtration system. In first
to the NO− 3 -N or free nitrogenous forms. few days, ammonia got accumulated in the reactor due to ammonifica-
The present systems consist of two units: VF unit favours ammonifica- tion of organic nitrogen and less development of ammonia oxidizing
tion and nitrification processes, and HF unit favours denitrification pro- bacteria (AOB). This is followed by a fall in pH in-between 6.9 and 7.1
cess. Diffused atmospheric oxygen gets transferred to the filter bed up due to conversion of ammonia into nitrate. Formation of organic acids
to few centimeters down from the surface of VF and HF unit. According and carbon dioxide also speed up the decrease of pH. The collective in-
to the two film theory, oxygen gas becomes transferred to wastewater fluence of these two processes controls the pH of all the reactors and
during the flow to the filter bed. In VF unit, wastewater flows from top lead towards neutrality (Samal et al., 2017b).
to bottom zone, and as it flows downward, oxygen is being consumed
and transferred to the filter bed converting it to aerobic in nature, 3.3. RSM model design for COD removal
which facilitate ammonification and nitrification process. Organic N
−
gets converted to NH+ +
4 -N and then to NO3 -N, whereas NH4 -N directly Interactive effects of different variables (% COD removal) were ana-
oxidised to NO− 3 -N in the presence of O2 . Different sources of O2 in reac- lyzed by performing BBD based experimental design. The predicted re-
tors is described in Fig. 4. Effluent from the VF unit channelled to the HF sponse showed a wide range of COD removal (63.89–82.5%), indicating
unit and there it flows horizontally at the bottom of the HF reactor. that the performance of the reactor is strongly dependent on variables.
Pore spaces of bottom zone of HF unit are filled with water and trans- A quadratic model is derived by the RSM design for % COD removal,
forms it to the anoxic/anaerobic region, which aid denitrification process which is given by three coded values in below Eq. (17).
(Samal et al., 2018). Combination of nitrification-denitrification process
remove major portion of nitrogen from wastewater. Plant root system COD removal ð%Þ ¼ þ82:50−4:08A−5:04B þ 1:76C−1:82AB
also support the nourishment of nitrifying and ammonium oxidizing bac- þ 0:93 AC
teria (AOB). Moreover, a lot of nitrifying and denitrifying bacteria are also þ 0:30 BCE−3:45A2 −4:22B2 −2:78C2 ð17Þ
present in earthworm's intestine. Plants uptake nitrogen from the water
and soil in the form of nitrate (NO− +
3 ) and ammonium (NH4 ). Ammonium
where, A, B, C are the coded terms for the variables as stated in Table 3.
ions are absorbed by the plant via ammonia transporters. Nitrate is taken
The positive sign (+) indicates the synergistic effect and negative sign
up by several nitrate transporters that use a proton gradient to power the
(−) indicates antagonistic effects for each term. Eq. (17) was used to
transport. Nitrogen is transported from the root to the shoot via the
calculate the predicted value of percentage COD removal and listed in
xylem in the form of nitrate, dissolved ammonia (Vymazal, 2007).
Table 4. The experimental values were fitting perfectly to the predicted
Fig. 6a shows NH+ 4 -N concentration in effluent of active layer and dif-
values and ANOVA was used to find the significance of the model.
ferent reactor units. In the present study, wastewater contain nitrogen in
Table 5 represents the ANOVA for COD removal quadratic model.
organic as well as NH+ 4 -N form. Simultaneously, organic N gets converted
Regression model significant is vindicated by the larger F-values, small
to NH+ +
4 -N and original NH4 -N present in wastewater is nitrified in oxy-
p-values, non-significant lack of fit and high values of R2 (Debnath and
genated condition. The DO profile of effluent from VF and HF unit is pre-
Gupta, 2018). p-values b0.0500 labeled high significant regression at
sented in Fig. 6b and the effluent DO in VF unit was found to be in
95% confidence level. The quadratic model for COD removal was
between 3.2 and 4.2 mg L−1. Ammonification and nitrification typically
occurred in active layer of VF unit and it depends on active layer depth, in-
fluent nitrogen concentration and HLR. Effluent from active layer Table 4
−1
contained NH+ 4 -N in the range of 11.5–29.7 mg L . In experimental Experimental design for optimization of COD removal.
run 11, effluent from active layer contained NH4 -N 11.5 mg L−1, while in-
+
Standard Actual variables Removal efficiency (%)
fluent concentration was 21.2 mg L−1. In most of the experimental run (1, order
HLR COD Active layer Experimental Predicted
2, 5, 7, 8, 9, 11, 12, 13, 14, 15) the NH+ 4 -N content in VF effluent was below
(m/d) (mg L−1) depth (cm)
8.5 mg L−1 and it is due to the oxidation of NH+ 4 -N in the VF unit. Remain-
ing amount of NH+ 1 0.30 1200.00 30.00 82.20 82.11
4 -N in the effluent of VF unit was up taken by macro-
2 0.90 1200.00 30.00 77.70 77.61
phyte in HF unit of the reactor and very limited nitrification may 3 0.30 2800.00 30.00 75.60 75.69
happens there. The average DO amount in the effluent of HF unit was 4 0.90 2800.00 30.00 63.80 63.89
1.43 mg L−1. In experimental run 1, 9, 13 and 14, concentration of 5 0.30 2000.00 15.00 79.60 79.51
NH+ 4 -N in HF effluent were 2.5, 2.4, 2.3 and 2.6, respectively, which
6 0.90 2000.00 15.00 69.60 69.51
7 0.30 2000.00 45.00 81.10 81.19
were very low. Simultaneously, increase of NO− 3 -N concentration in the
8 0.90 2000.00 45.00 74.80 74.89
VF effluent, indicated the establishment of aerobic condition in VF unit 9 0.60 1200.00 15.00 78.90 79.08
of the reactor. The average DO amount in the effluent of VF unit was 10 0.60 2800.00 15.00 68.40 68.40
3.83 mg L−1. In case of experimental run 3, 10 and 12, NO− 3 -N concentra- 11 0.60 1200.00 45.00 82.00 82.00
tion in VF unit was higher than 80 mg L−1 (Fig. 6c). Influent TN concentra- 12 0.60 2800.00 45.00 72.70 72.53
13 0.60 2000.00 30.00 82.40 82.50
tion in all of the above experimental runs were 148 mg L−1 and DO 14 0.60 2000.00 30.00 83.20 82.50
concentration in VF unit were 3.9, 3.8 and 4 mg L−1 (Experimental run 15 0.60 2000.00 30.00 81.90 82.50
3, 10, 12, respectively). At high TN influent concentration and DO, activity
164 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169

Table 5 Table 6
ANOVA for response surface quadratic model of percentage COD removal. Experimental design for optimization of TN removal.

Source Sum of Degrees of Mean F value p-value Standard Actual variables Removal efficiency (%)
squares freedom square prob N F order
HLR TN Active layer Experimental Predicted
Model 498.51 9 55.39 281.88 b0.0001 Significant (m/d) (mg L−1) depth (cm)
A-HLR 132.85 1 132.85 676.06 b0.0001
1 0.30 64.00 30.00 55.80 56.06
B-COD 203.01 1 203.01 1033.14 b0.0001
2 0.90 64.00 30.00 43.40 43.46
C-active 24.85 1 24.85 126.47 b0.0001
3 0.30 148.00 30.00 41.50 41.44
layer depth
4 0.90 106.00 30.00 31.80 31.54
AB 13.32 1 13.32 67.80 0.0004
5 0.30 106.00 15.00 47.70 47.16
AC 3.42 1 3.42 17.42 0.0087
6 0.90 106.00 15.00 35.90 35.56
BC 0.36 1 0.36 1.83 0.2339
7 0.30 106.00 45.00 51.20 51.54
A2 43.95 1 43.95 223.65 b0.0001
8 0.90 106.00 45.00 40.10 40.64
B2 65.91 1 65.91 335.42 b0.0001
9 0.60 64.00 15.00 44.90 45.18
C2 28.43 1 28.43 144.70 b0.0001
10 0.60 148.00 15.00 34.00 34.60
Residual 0.98 5 0.20
11 0.60 64.00 45.00 53.20 52.60
Lack of fit 0.12 3 0.041 0.095 0.9560 Not
12 0.60 148.00 45.00 36.90 36.63
significant
13 0.60 106.00 30.00 56.90 56.63
Pure error 0.86 2 0.43
14 0.60 106.00 30.00 57.30 56.63
Cor total 499.49 14
15 0.60 106.00 30.00 55.70 56.63
Model
summary
statistics
Source Std. R2 Adjusted Predicted Press (Table 5). Fig. 7b–d shows the 3D response surface plots of HLR (m/d),
Dev. R2 R2
COD (mg L−1) and active layer depth (cm). From Fig. 7b, it is clearly
Linear 3.55 0.7222 0.6464 0.5921 203.77
2FI 3.90 0.7564 0.5737 0.4893 255.11 depicted that when HLR value increases beyond 0.6 m/d, COD re-
Quadratic 0.44 0.9980 0.9945 0.9922 3.90 Suggested moval efficiency started decreasing and similarly, for influent COD
Cubic 0.66 0.9983 0.9879 Aliased concentration of 2800 mg L−1, removal efficiency was least. An ac-
tive layer depth of 30 cm was optimum for pollutants degradation
process (Fig. 7c). The retention of pollutants in filter bed was less
statistically significant with high F-value-281.88, low p-value- b 0.0001, at high hydraulic load. So, the interaction duration among microbes,
non-significant lack of fit 0.12, R2 0.9980, adjusted R2 0.9945 and pre- pollutants and earthworms shortens, which reduce biodegradation
dicted R2 0.9922. The relation between the actual and predicted values of organic pollutants (Samal et al., 2018). In VF unit of the system
of COD removal exhibits an excellent agreement (Fig. 7a). major portion of the pollutants filtered out and deposited on the
top zone of the earthworm active layer, which was digested by earth-
3.4. Variables interaction and the effect on response worms. When depth of the worm active zone (soil layer) is higher,
distribution of earthworm population may not be uniform through-
3D response surface plots was used to analyze the interaction effects out that layer. There is chance of migration of worms to the top
of independent variables for COD removal response. According to ‘F’ and zone of active layer due to availability of oxygen. In general, when
‘p’ values, the terms A, B, C, AB, AC, A2, B2 and C2 were significant while density of earthworms becomes higher than a certain limit in a par-
BC interaction effect was found non-significant at a level of 0.05 ticular area, it negatively affects earthworm growth and metabolism

(a)

Removal efficiency (%)

Predicted

(b)

A: HLR
B: COD

Actual

Removal efficiency (%) Removal efficiency (%)

(c)
(d)

C: Active layer depth A: HLR

C: Active layer depth B: COD

Fig. 7. 3-D Response surface plot showing interactive effect of (a) Correlation of actual vs. predicted COD removal efficiency, (b) effect of HLR and COD of wastewater, (c) effect of HLR and
active layer depth, (d) effect of COD of wastewater and active layer depth.
 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169 165

(a)

Removal efficiency (%)

Predicted

(b)

X: HLR

Y: TN concentration

Removal efficiency (%) Removal efficiency (%)

(c) (d)

Y: TN concentration
Z: Active layer depth X: HLR Z: Active layer depth

Fig. 8. 3-D Response surface plot showing interactive effect of (a) Correlation of actual vs. predicted TN removal efficiency, (b) effect of HLR and TN concentration of wastewater, (c) effect
of HLR and active layer depth, (d) effect of TN concentration of wastewater and active layer depth.

process, which ultimately reduce treatment practice (Wang et al.,
2013; Xu et al., 2013). There should be a balance between active
layer depth and hydraulic load of wastewater in MAVF. In 15 cm ac-
tive layer depth, retention time was insufficient for degradation of
pollutants. The experimental COD removal efficiency was varied
from 63.8% to 83.2%. The highest COD removal efficiency (83.2%)
was observed at HLR 0.6 m/d, COD 2000 mg L−1 and active layer
depth 30 cm.
where, X, Y, Z are the coded terms for the variables as stated in Table 3.
Table 6 represents the ANOVA for response surface quadratic model of
TN removal. It was statistically significant with high F-value-191.77,
low p-value- b 0.0001, non-significant lack of fit 1.82, R2 0.9971, ad-
justed R2 0.9919 and predicted R2 0.9709. The actual and predicted
values of TN removal exhibits good correlation (Fig. 8a).
3.6. Variables interaction and the effect on response

3.5. RSM model design for TN removal TN removal was also investigated using 3D response surface plots
and by analyzing the ‘F' and ‘p’ values from Table 7, it was found that
Like COD removal, a quadratic model is developed for TN removal the terms X, Y, Z, YZ, ×2, Y2 and Z2 were significant at a level of 0.05.
(%) and is given by Eq. (18). The predicted TN removal response varied The 3D response surface plots of HLR (m/d), COD (mg L−1) and active
from 31.54% to 56.63%. layer depth (cm) are shown in Fig. 8b–d. As influent TN concentration
increased from 64 to 106 mg L−1, removal efficiency also increased. It
may be due to the abundance of ammonium oxidizing bacteria (AOB),
TN removal ð%Þ ¼ þ56:63−5:63  −6:64Y þ 2:36Z þ 0:68XY nitrifying and denitrifying bacteria. Due to the availability of more sub-
þ 0:18XZ−1:35YZ−6:02  2 −7:49Y2 −6:89Z2 ð18Þ strates, microbes consume and convert most of the influent nitrogenous

Table 7
ANOVA for response surface quadratic model of percentage TN removal.

Source Sum of squares Degrees of freedom Mean square F value p-value prob N F

Model 1107.77 9 123.09 191.77 b0.0001 Significant

A-HLR 253.13 1 253.13 394.38 b0.0001
B-TN concentration 352.45 1 352.45 549.13 b0.0001
C-active layer depth 44.65 1 44.65 69.57 0.0004
AB 1.82 1 1.82 2.84 0.1528
AC 0.12 1 0.12 0.19 0.6804
BC 7.29 1 7.29 11.36 0.0199
A2 133.66 1 133.66 208.25 b0.0001
B2 207.23 1 207.23 322.87 b0.0001
C2 175.37 1 175.37 273.23 b0.0001
Residual 3.21 5 0.64
Lack of fit 1.82 3 0.61 0.88 0.5720 Not significant
Pure error 1.39 2 0.69
Cor total 1110.98 14
Model summary statistics
Source Std. Dev. R2 Adjusted R2 Predicted R2 Press
Linear 6.47 0.5853 0.4722 0.4307 632.50
2FI 7.51 0.5936 0.2888 0.1762 915.20
Quadratic 0.80 0.9971 0.9919 0.9709 32.28 Suggested
Cubic 0.83 0.9988 0.9913 Aliased
166 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169
substances, and their body metabolism rate also increases. Similarly,
when HLR increased from 0.3 to 0.9 m/d, due to the decrease in contact
time, conversion of nitrogenous substances slows down and TN concen-
tration in effluent increases (Ubay, 1994; Ramakrishnan and Gupta,
2008). Active layer depth has a major impact on TN removal. As the
depth of soil layer increases, stocking density of earthworm also in-
crease in reactor, which help in conversion and degradation of nitroge-
nous substances. Due to the earthworm burrowing activity,
atmospheric oxygen diffused to the filter bed and aerobic condition
maintained up to a certain limit, which accelerate nitrification process.
When active layer depth increase from 15 to 30 cm in VF unit, wastewa-
ter flow path rises and contact time between microbes and pollutants
too, causing more removal of TN from influent wastewater (Samal
et al., 2017b; Singh et al., 2017; Kantawanichkul et al., 2009). Experi-
mental TN removal efficiency varied from 31.8% to 57.3% in MAVF
Fig. 9. (a) First-order model plot for COD, (b) Grau second-order model plot for COD and (c) TN, (d) Stover-Kincannon model plot for COD and (e) TN.
reactor. The highest TN removal efficiency was observed at HLR
0.6 m/d, TN 106 mg L−1 and active layer depth 30 cm.
3.7. Optimization of process parameters and model validation
Optimization was done by applying a multiple response method,
called desirability function (D). The individual desirability (di) of each
response and variable was set from 0 and 1 and their desirability score
was combined into overall desirability function as follows:
1
Yn 1n
D ¼ ðd1  d2  d3 Þn ¼ d
i¼1 i
ð19Þ
The simultaneous optimization of COD and TN removal efficiency by
desirability functions was carried out with Design Expert 7.0.0. In point
 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169 167

optimization of COD removal, maximum removal efficiency, maximum 3.8.2. Grau second-order model
HLR, maximum wastewater strength, in range active layer depth in VF The second-order model was also used to evaluate the rate of the
were selected as the most desirable targeted criteria for the optimum biodegradation process occurred in the MAVF systems. In order to find
response. Out of all alternative solutions, the maximum desirability out the kinetic coefficients (a, b and K2), Fig. 9b and c was plotted as
was found to be a removal efficiency of 82.2% for COD at HLR per the Eq. (6), and the values of kinetic coefficients were calculated
0.63 m/d, COD concentration 2015.5 mg L−1 and active layer depth from the intercept and the slope of straight line on the graph. The values
33.5 cm. TN removal efficiency was 55.6% with high desirability value of ‘a’ and ‘b’ were found to be 0.0977 and 1.195, respectively for COD and
at similar condition. 0.4043 and 1.8026, respectively for TN. The correlation coefficient (R2)
The values for COD and TN removal efficiency were predicted using for COD and TN were 0.9817 and 0.8395, respectively. Effluent substrate
the developed models, i.e., Eqs. (17) and (18) respectively, for various concentration can be determined by rearranging and substituting the
experimental conditions from run 1 to 15. Figs. 7a and 8a represent value of ‘a’ and ‘b’ in Eq. (8). Based on the Eq. (8), the relationship be-
the plot of observed and predicted removal efficiency of COD and TN, re- tween effluent pollutants concentration and HRT is described as fol-
spectively. The correlation between the observed and predicted values lows:
demonstrates that the predicted values by the developed model were

reasonably close to the experimental value indicating the validity of ƟH
For COD; Se ¼ Si 1− ð20Þ
the predicted model. Therefore, the developed quadratic models can 0:0977 þ 1:195 ƟH
be used to design required efficiency within the experimental

conditions. ƟH
For TN; Se ¼ Si 1− ð21Þ
0:4043 þ 1:8026 ƟH

3.8. Kinetic modelling
3.8.1. First-order model
The majority of biological wastewater treatment processes are
described by first-order kinetics. Reaction orders can differ when
there is variation in the microorganisms, substrates and environ-
mental conditions. The organic pollutants removal has been tradi-
tionally modeled as a continuous first order reaction, in which the
rate of breakdown is at first rapid, but gets progressively slower as
the organic material is utilized (Priya et al., 2009; Kaewsuk et al.,
2010; Sponza and Uluko, 2008).
The first-order kinetic constant K1 was calculated by plotting (Si–Se)/
ƟH versus Se for COD. From Fig. 9a, the kinetic coefficient (K1) was deter-
mined to be 2.0314 d−1 with R2 of 0.5212. The model did not show a
good agreement with the data obtained for TN. Mansouri et al.
(Mansouri et al., 2014) obtained kinetic coefficient of 14.549 d−1 for
COD while treating synthetic domestic wastewater in an up-flow anoxic
sludge fixed film bioreactor. Akhbari et al. (Akhbari et al., 2012) re-
ported 1.882 d−1 for TN removal in a rotating biological contactor-
activated sludge system. Table 8 shows the kinetic coefficient values
for different models.
The average value of K2 was calculated as 0.16 d−1 using the equa-
tion: a = S0/(K2.X), which reflects the substrate removal rate for each
unit of microorganism. Ubay (Ubay, 1994) obtained the values of a, b
and K2 for COD as 0.002, 1.346 and 0.217 d−1, respectively; for the treat-
ment of municipal wastewater in UASB. Isik and Sponza (Isik and
Sponza, 2005) presented the values of kinetic coefficients a, b and K2
as 0.562, 1.095 and 0.337 d−1, respectively for synthetic textile waste-
water treatment. Optaken et al. (Optaken, 1982) reported ‘a’ and ‘b’
value as 0.033 and 1.192 and K2 value 10.81 d−1; while treating
molasses.
3.8.3. Stover-Kincannon model
Stover-Kincannon model was applied for kinetic modelling of COD
and TN removal in the MAVF system. Fig. 9d and e was plotted as per
the Eq. (13) between reciprocal of organic loading removal rate and re-
ciprocal of the total organic loading rate for determining the values of
maximum utilization rate (Umax) and saturation value constant (Ks). In-
tercept and slope was determined by using linear regression methods.
Ks and Umax for COD removal was 20.4 and 17.39 g L−1 d−1, respectively
and for TN; 0.2862 and 0.1771 g L−1 d−1, respectively. The correlation
coefficients (R2) for COD, N removal were 0.9961, 0.9353, respectively.

Table 8
Comparison of kinetic coefficients for different models.

Models Substrate Type of COD in (g HRT (d) Kinetic References

reactors L−1) parameters

K1 (d−1)
First-order Synthetic wastewater UAASFF 1 0.083–0.271 12.09–30.71 (Mansouri et al., 2014)
Dairy wastewater MAVF 1.2–2.8 0.53–1.79 2.0314 Present study

K2 (d−1) a b
Grau second-order Molasses RBC 2–15 0.5–2.0 10.81 0.033 1.192 (Optaken, 1982)
Synthetic wastewater UAASFF 1 0.083–0.271 5.95 0.042 0.928 (Mansouri et al., 2014)
UAIB 0.75–4.5 0.5–1.0 3.582 0.047 1.007 (Borghei et al., 2008)
Synthetic coal wastewater Hybrid UASB 0.75–1.5 1.72 0.078 0.964 (Ramakrishnan and Gupta, 2008)
Formaldehyde-containing UAFB 0.42–1 0.133 0.64 9.36 (Priya et al., 2009)
wastewater
Dairy wastewater MAVF 1.2–2.8 0.53–1.79 0.16 0.0977 1.195 Present study

Umax (L−1 d−1) Ks (L−1 d−1)

Stover-kincannon Synthetic wastewater MBBR 0.75–4.5 1 8.3 9.45 (Borghei and Hosseiny, 2002)
Soybean wastewater AF 7.5–11.45 1–1.45 83.3 85.5 (Kaewsuk et al., 2010)
Synthetic wastewater UAASFF 1 0.083–0.271 38.46 37.88 (Mansouri et al., 2014)
Simulated textile wastewater 4.21 7.501 8.211 (Pirsaheb et al., 2009)
Poultry slaughterhouse 1.6–9.1 12.148 130.28 (Büyükkamac and Filibeli, 2002)
Molasses 2.0–15.0 83.3 186.23 (Yu et al., 1998)
Fruit canning wastewater UAF 0.5 109.9 53.5 (Rajagopal et al., 2013)
Dairy wastewater MAVF 1.2–2.8 0.53–1.79 17.39 20.4 Present study
168 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169

It confirms the suitability of the Stover-Kincannon model. Effluent con- components and recalcitrant compounds in wastewater causes low
centration can be calculated from substrate mass balance equation as constant value. Borghei et al. (Borghei et al., 2008) found a high Umax
follows: value (COD 101 g L−1 d−1), whereas Borghei and Hosseiny (Borghei
and Hosseiny, 2002) found a low Umax value (COD 8.3 g L−1 d−1). Sim-
For COD; Se ¼ Si −
17:39 Si

 ð22Þ ilarly, Umax value for TN removal (0.1771 g L−1 d−1) was lower than the
20:4 þ
Q Si values found by Jin and Zheng (Jin and Zheng, 2009) (12.4 g L−1 d−1),
V which may be due to high nitrogen loading rate. Umax depends on vari-
ous factors like OLR, characteristics of wastewater, types of reactor and
0:1771 Si microorganisms used.
For TN; Se ¼ Si −
 ð23Þ
Q Si
0:2862 þ
V 4. Earthworms and macrophytes growth characteristics

The calculated values indicate the substrate removed by the earth-
worms and microorganisms during the given time and thereby reflects
the biodegradability of the organic matter under the current operating
conditions. The Stover-Kincannon model could be used in the design
of the MAVF reactor. In Table 8, a wide range of kinetic constants
(Umax and Ks) have been reported for various types of wastewater and
experimental conditions. Easily biodegradable substrate like glucose
and molasses causes high constant value, whereas various inorganic
Fig. 10 represents the earthworm growth characteristics in the ex-
periment. Percentage increase in number and biomass of earthworm
in all VF unit of the reactor were in the range of 26.5–32.4 and
30.2–33.9, respectively and for HF unit were 19.2–25.5 and 21.8–29.7,
respectively. It was observed that the root system of Canna indica pene-
trated in to the gravel media of the filter making bottom zone
favourable for earthworm occupancy. So, some worm drifted to the bot-
tom zone of the filter bed. The coarse and irregular surface on gravel and

Fig. 10. Earthworm growth characteristics: (a) % increase in number (b) % increase in biomass.
 K. Samal et al. / Science of the Total Environment 645 (2018) 156–169
sand resulted scratches on the body of the earthworms triggering their
death (Samal et al., 2017b). HLR and wastewater strength has no signif-
icance impact on earthworm growth.
Canna indica plant had a healthy growth during the entire experi-
mental period. Initially one plant planted in VF unit and at the end it is
split in to 5 to 8 plants in all reactors and similar case also observed in
HF unit of the reactor. Above ground part of Canna indica species grew
up to a height of 1.2 to 1.6 m and root system of all macrophytes reached
up to the bottom of the VF reactor, which were 55–85 cm depth. Plant
roots maintain constant infiltration rate by penetrating in the filter
bed, so prevent clogging process and maximum root length of the
plant is useful in calculating the suitability for MAVF.
5. Conclusions
Various parameters of hybrid MAVF (HLR, wastewater strength and
depth of active layer in VF unit) have been optimized using response
surface methodology (RSM) for COD and TN removal. 83.2% COD and
57.3% TN were removed in the MAVF system at an active layer depth
of 30 cm and HLR of 0.6 m/d. The model was validated by fitting the
values of the variables to the developed model, and the results (pre-
dicted values) were in good agreement with experimental values.
From the kinetic modelling analysis, it was observed that Stover-
Kincannon model fitted well to the degradation of all the pollutants.
The characteristics of earthworms and plant growth were also analyzed
at the end of the experiments; but it can be concluded that HLR and
wastewater strength have no major impact on both earthworm popula-
tion and biomass growth. VF unit of MAVF ensured the nitrification pro-
cess and HF unit promoted the denitrification process. The combined
process removed TN from wastewater effectively. Overall, MAVF system
worked well for effective reduction of pollutants in dairy wastewater.
Acknowledgements
The authors wish to thank the Department of Civil Engineering,
School of Infrastructure, Indian Institute of Technology Bhubaneswar,
India, for providing facilities for carrying out research work in the re-
lated area.
References
Akhbari, A., Zinatizadeh, A.A.L., Mohammadi, P., Mansouri, Y., Irandoust, M., Isa, M.H.,
2012. Kinetic modeling of carbon and nutrients removal in an integrated rotating bi-
ological contactor-activated sludge system. Int. J. Environ. Sci. Technol. 9, 371–378.
APHA (Ed.), 2005. Standard Methods for the Examinations of Water and Wastewater,
twenty-first ed. APHA and AWWA and WEF, DC, Washington.
Arias, C.A., Brix, H., Marti, E., 2005. Recycling of treated effluents enhances removal of total
nitrogen in vertical flow constructed wetlands. J. Environ. Sci. Health, Part A: Tox.
Hazard. Subst. Environ. Eng. 40, 1431–1443.
Arora, S., Rajpal, A., Kazmi, A.A., 2016. Antimicrobial activity of bacterial community for
removal of pathogens during vermifiltration. J. Environ. Eng. 142, 1–10.
Bajsa, O., Nair, J., Mathew, K., Ho, G.E., 2003. Vermiculture as a tool for domestic wastewa-
ter management. Water Sci. Technol. 48, 125–132.
Borghei, S.M., Hosseiny, S.H., 2002. Modeling of organic removal in a moving bed biofilm
reactor (MBBR). Sci. Iran. 9, 53–58.
Borghei, S., Sharbatmaleki, M., Pourrezaie, P., Borghei, G., 2008. Kinetics of organic re-
moval in fixed-bed aerobic biological reactor. Bioresour. Technol. 99, 1118–1124.
Büyükkamac, N., Filibeli, A., 2002. Determination of kinetic constants of an anaerobic hy-
brid reactor. Process Biochem. 38, 73–79.
Carballeira, T., Ruiz, I., Soto, M., 2017. Aerobic and anaerobic biodegradability of accumu-
lated solids in horizontal subsurface flow constructed wetlands. Int. Biodeterior.
Biodegrad. 119, 396–404.
Dan, T.H., Quang, L.N., Chiem, N.H., Brix, H., 2011. Treatment of high-strength wastewater
in tropical constructed wetlands planted with Sesbania sesban: horizontal subsurface
flow versus vertical downflow. Ecol. Eng. 37, 711–720.
169
Debnath, D., Gupta, A.K., 2018. Optimizing the fabrication of nano-plasmonic silver nitro-
gen co-doped zinc oxide (AgxZn(1-x)NyO(1-y)) mediated by ammonia template: in-
sight into its enhanced physiochemical and photocatalytic behaviour. J. Mol. Liq. 249,
334–345.
Demirel, B., Yenigun, O., Turgut, T.O., 2005. Anaerobic treatment of dairy wastewaters: a
review. Process Biochem. 40, 2583–2595.
Hait, S., Tare, V., 2011. Vermistabilization of primary sewage sludge. Bioresour. Technol.
102, 2812–2820.
Isik, M., Sponza, D.T., 2005. Substrate removal kinetics in an upflow anaerobic sludge
blanket reactor decolorising simulated textile wastewater. Process Biochem. 40,
1189–1198.
Jin, R.C., Zheng, P., 2009. Kinetics of nitrogen removal in high rate anammox upflow filter.
J. Hazard. Mater. 170, 652–656.
Kaewsuk, J., Thorasampan, W., Thanuttamavong, M., Seo, G.T., 2010. Kinetic development
and evaluation of membrane sequencing batch reactor (MSBR) with mixed cultures
photosynthetic bacteria for dairy wastewater treatment. J. Environ. Manag. 91,
1161–1168.
Kantawanichkul, S., Kladprasert, S., Brix, H., 2009. Treatment of high-strength wastewater
in tropical vertical flow constructed wetlands planted with Typha angustifolia and
Cyperus involucratus. Ecol. Eng. 35, 238–247.
Liu, J., Lu, Z., Yang, J., Xing, M., Yu, F., Guo, M., 2012. Effect of earthworms on the perfor-
mance and microbial communities of excess sludge treatment process in vermifilter.
Bioresour. Technol. 117, 214–221.
Mansouri, A.M., Zinatizadeh, A.A.L., Akhbari, A., 2014. Kinetic evaluation of simultaneous
CNP removal in an up-flow aerobic/anoxic sludge fixed film (UAASFF) bioreactor.
Iran. J. Energ. Environ. 5, 323–336.
Optaken, E.J., 1982. Rotating biological contactor-second order kinetics. Proceedings of
the 1st International Conference on Fixed Film Biological Processes. Kings Island,
Ohio.
Pirsaheb, M., Mesdaghin, A.R., Shahtaheri, S.J., Zinatizadeh, A.A., 2009. Kinetic evaluation
and process performance of a fixed film bioreactor removing phthalic acid and di-
methyl phthalate. J. Hazard. Mater. 167, 500–506.
Priya, K.R., Sandhya, S., Swaminathan, K., 2009. Kinetic analysis of treatment of formalde-
hyde containing wastewater in UAFB reactor. Chem. Eng. J. 148, 212–216.
Rajagopal, R., Torrijos, M., Kumar, P., Mehrotra, I., 2013. Substrate removal kinetics in
high-rate upflow anaerobic filters packed with low-density polyethylene media
treating high-strength agro-food wastewaters. J. Environ. Manag. 116, 101–106.
Ramakrishnan, A., Gupta, S.K., 2008. Effect of hydraulic retention time on the biodegrada-
tion of complex phenolic mixture from simulated coal wastewater in hybrid UASB re-
actors. J. Hazard. Mater. 153, 843–851.
Rodgers, M., Xiao, L.W., Mulqueen, J., 2006. Synthetic dairy wastewater treatment using a
new horizontal-flow biofilm reactor. J. Environ. Sci. Health A 41, 751–761.
Samal, K., Dash, R.R., Bhunia, P., 2017a. Treatment of wastewater by vermifiltration inte-
grated with macrophyte filter: a review. J. Environ. Chemi. Eng. 5, 2274–2289.
Samal, K., Dash, R.R., Bhunia, P., 2017b. Performance assessment of a Canna indica assisted
vermifilter for synthetic dairy wastewater treatment. Process. Saf. Environ. Prot. 111,
363–374.
Samal, K., Dash, R.R., Bhunia, P., 2018. Effect of hydraulic loading rate and pollutants deg-
radation kinetics in two stage hybrid macrophyte assisted vermifiltration system.
Biochem. Eng. J. 132, 47–59.
Singh, R., Bhunia, P., Dash, R.R., 2017. A mechanistic review on vermifiltration of waste-
water: design, operation and performance. J. Environ. Manag. 197, 656–672.
Sinha, R.K., Bharambe, G., Chaudhari, U., 2008. Sewage treatment by vermifiltration with
synchronous treatment of sludge by earthworms: a low-cost sustainable technology
over conventional systems with potential for decentralization. Environment 28,
409–420.
Sponza, D.T., Uluko, A., 2008. Kinetic of carbonaceous substrate in an upflow anaerobic
sludge sludge blanket (UASB) reactor treating 2,4 dichlorophenol (2,4 DCP).
J. Environ. Manag. 86, 121–131.
Ubay, E., 1994. Anaerobic treatment of municipal wastewater. Dissertation. Faculty of Sci-
ence, Istanbul Technical University.
Vymazal, J., 2007. Removal of nutrients in various types of constructed wetlands. Sci. Total
Environ. 380, 48–65.
Wang, L., Luo, X., Zhang, Y., Chao, J., Gao, Y., Zhang, J., Zheng, Z., 2013. Community analysis
of ammonia-oxidizing betaproteobacteria at different seasons in microbial-
earthworm ecofilters. Ecol. Eng. 51, 1–9.
Xing, M., Li, X., Yang, J., 2010. Treatment performance of small-scale vermifilter for do-
mestic wastewater and its relationship to earthworm growth, reproduction and en-
zymatic activity. Afr. J. Biotechnol. 9, 7513–7520.
Xu, D., Li, Y., Howard, A., 2013. Influence of earthworm Eisenia fetida on removal efficiency
of N and P in vertical flow constructed wetland. Environ. Sci. Pollut. Res. 20,
5922–5929.
Yu, H.Q., Wilson, F., Tay, J.H., 1998. Kinetic analysis of an anaerobic filter treating soybean
wastewater. Water Res. 32, 3341–3352.