Applied Energy 194 (2017) 287–295

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Three-stage anaerobic digester for food waste

Jingxin Zhang a, Kai-Chee Loh b, Wangliang Li a, Jun Wei Lim a, Yanjun Dai c, Yen Wah Tong a,b,⇑
a
Environmental Research Institute, National University of Singapore, Singapore
b
Department of Chemical and Biomolecular Engineering, NUS, Singapore
c
School of Mechanical Engineering, Shanghai Jiao Tong University, China

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

A compact three-stage anaerobic

digester (TSAD) was developed for
food waste treatment.

TSAD combined the advantages of
high solids AD and Wet AD.

Methane yields in TSAD was
increased by 24–54% compared to
traditional anaerobic digesters.

Functionalized partitioning in TASD
significantly improved hydrolysis/
acidogenesis efficiency.

TASD has higher treatment capacity
and solid reduction rate with less
reactor volume requirement.

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

Article history: A novel compact three-stage anaerobic digester (TSAD) was developed for high-efficiency anaerobic
Received 26 July 2016 digestion of food waste and methane production. Through structure optimization by having three sepa-
Received in revised form 4 October 2016 rate chambers in a single-stage anaerobic digester, hydrolysis, acidogenesis and methanogenesis were
Accepted 27 October 2016
independently optimized with concomitant improvement in anaerobic digestion performance.
Available online 8 November 2016
Compared to traditional one-stage and two-stage anaerobic digesters, TSAD had a 24–54% higher
methane yield at a high organic loading rate of 10 g VSFW/L. A higher volatile solid reduction rate of
Keywords:
83.5 ± 1.3% was also achieved in TSAD. Even at high organic loading, TSAD still presented a high buffering
Anaerobic
Three-stage anaerobic digestion
ability when the one-stage and two-stage digesters had already soured and failed. Pyrosequencing
Food waste analysis indicated that the bacterial community in TSAD is more diverse than the control digesters.
Microbial community Multi-function methanogens Methanosarcina and some dominant populations with the function of
acetogenesis, amino-acid-utilization and symbiosis were found to selectively enrich in TSAD.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction

Research on sustainable and renewable bioenergy derived from

⇑ Corresponding author at: Department of Chemical and Biomolecular Engineer-
organic solid wastes is increasingly important and popular because
ing, National University of Singapore, 4 Engineering Drive 4, Singapore 117585,
Singapore.
of increasing global population, depletion of natural energy
E-mail address: chetyw@nus.edu.sg (Y.W. Tong). resources, and the environmental burden brought on by the

http://dx.doi.org/10.1016/j.apenergy.2016.10.116
0306-2619/Ó 2016 Elsevier Ltd. All rights reserved.
288 J. Zhang et al. / Applied Energy 194 (2017) 287–295
generation of organic solid wastes [1,2]. Among the many waste-
to-bioenergy technologies being explored, anaerobic digestion
(AD) is commonly used for organic waste treatment, renewable
energy production and the recycling of nutrients [3,4]. Through
AD, biodegradation of organic wastes generates biogas (55–65%
CH4; 35–45% CO2) that can be converted to electricity and/or heat
[5–7]. To a limited extent, AD bioenergy can be a good supplement
to conventional fossil energy, as well as mitigate greenhouse gas
emissions.
On the basis of total solids (TS) content, AD can be categorized
into wet AD (TS < 15%) or high-solids AD (15% < TS < 40%) [8–10].
Wet AD is more widely applied in practice as the high moisture
content promotes the growth of the microorganisms through
enhancement of mass transfer between substrates and microor-
ganisms [11,12]. However, wet AD suffers from the need for larger
reactor volume, high consumption of water, as well as expensive
post-treatment [9]. High-solids AD, on the other hand, is attracting
more attention because of its reduced capital and operating costs,
as well as its amenability to tackle a wide range of organic wastes
[10,13]. Nevertheless, high-solids AD also suffers from some tech-
nical bottlenecks such as high inoculum demand, long retention
time and accumulation of volatile fatty acids (VFAs) [14–17]. The
specific objective of this research was to design a novel anaerobic
digester combining the benefits of wet AD and high-solids AD to
enhance organic waste processing and methane production.
Given that the optimum pH for hydrolysis and acidogenesis
were 5.5 and 6.5 [9,18], respectively, and the suitable range of
pH for methanogenesis was 6.5–8.2 [19], a novel three-stage
anaerobic digester (TSAD) was developed and tested using food
waste (FW) as feedstock. Structurally, the TSAD is separated into
three vertical components i.e. high-solids hydrolysis, acidification
and wet methanogenesis, respectively. Based on this, it was antic-
ipated that the optimal condition of pH can be applied in each
stage, thereby improving overall efficiency. In addition, vertical
integration in the TSAD resulted in a small footprint, saving space
and created favorable environments for the growth of the different
functional microorganisms of hydrolyzing bacteria, fermenting
bacteria and methanogenic archaea.
Hitherto, most studies focus specifically on either wet AD or
high-solids AD [20–23], and little research has been conducted to
design and operate a compact AD reactor that combines both
wet AD and high-solids AD in one digester. In this study, the treat-
ment performance of TSAD and the synergistic effect of microor-
ganisms in the TSAD were investigated. Further to a proof-of-
concept for the TSAD operation on food waste, an in-depth study
of the microbial community structure using high-throughput 16S
rDNA pyrosequencing was also conducted.
2. Materials and methods
2.1. Inocula and substrates
Seed sludge was collected from a large-scale anaerobic digester
from the Ulu Pandan Water Reclamation Plant (UPWRP) in Singa-
pore. The anaerobic digester at UPWRP currently treats waste acti-
vated sludge from the secondary wastewater treatment plant for
domestic sewage wastes. In this study, each reactor was inoculated
with this seed sludge at approximately 80% (v/v). The ratio of vola-
tile suspended sludge (VS) to total suspended sludge was 0.65 with
initial TS of 11.8 g/L.
FW was obtained from a canteen at the National University of
Singapore. This comprised mainly rice, noodles, meat, vegetables,
and condiments. After removing any bones and non-
biodegradable waste like plastic bags and utensils, the FW was
homogenized by a blender and then stored at 20 °C freezer to
prevent biological decomposition. Detailed characteristics of FW
are listed in Table S1 (see Supplementary Material).
2.2. Reactor specifications and operation
Fig. 1 shows the schematic diagram of this TSAD system fabri-
cated for this research. Chambers level 1, level 2 and level 3 corre-
spond to high-solids hydrolysis stage, acidification stage and wet
methane-production stage, respectively. Food waste was first fed
into chamber level 1 for hydrolysis. After that, the hydrolyzed
FW was transferred to chamber level 2 for acidogenesis. Finally,
the hydrolyzed and acidified FW was dropped to chamber level 3
for methanogenesis. The baffle at the bottom of each chamber
can be opened by a connecting rod from the outside of the digester.
In this way, food waste was gravity transferred from one chamber
to another. The optimized pH in each chamber was adjusted and
controlled individually. To demonstrate the performance of the
TSAD, baseline studies involving 1 L AD reactors were carried out
to examine the individual stages of hydrolysis, acidogenesis and
methanogenesis, before fabricating a 20 L steel TSAD for proof-
of-concept. Essentially, three 1 L glass reactors (effective working
volume of 0.8 L) was set up in parallel, to simulate a one-stage
AD (SL1), a two-stage AD (SL2) and the TSAD. FW was adjusted
to 20% TS through the addition of seeding sludge and then stored
in the 20 °C freezer as feedstock for SL1. During the hydrolysis
stage, SL1 was operated at 35 °C with a stirrer speed of 150 rpm.
After 2 days of operation, the hydrolyzed FW was stored in the
freezer at 20 °C to be used as feedstock for SL2. For SL2 operation,
this hydrolyzed FW was mixed with seeding sludge to a TS of 10%.
pH was controlled at 6.5 ± 0.2, to provide the optimal conditions
for acidogenesis. After 2 days of operation, the acidified FW was
stored in the freezer at 20 °C to be used as feedstock for TSAD.
After adding the seed sludge, the three reactors (SL1, SL2 and
TSAD) were operated in a semi-continuous mode (feeding every
two or three days with the respective feedstocks) with gradual
increase in organic loading rates (OLR) of 1.6, 2.4, 4.0, 5.2, 8.0
and 10.0 g VSFW/L. All the experiments were conducted in tripli-
cates at the same experimental conditions.
After the baseline studies, a steel TSAD (diameter
250 mm ⁄ height 400 mm) was fabricated and operated. The inter-
nal structure of this TSAD system is shown in Fig. 1. Fresh FW was
added into the top chamber every day. The digestate was used as
inocula for hydrolysis and acidogenesis and recirculated from the
bottom of the TSAD to the top chamber; details of each step as
described above. The operating steps and conditions were the same
as that of the 1 L lab-scale ADs. The OLR was maintained at 2 g
VSFW/L.
2.3. Analytical methods
COD and ammonia were determined using HACH colorimeter
(HACH DR900, USA) according to the manufacturer’s instructions.
The pH was recorded using a pH analyzer (Agilent 3200 M, USA).
TS and VS were determined based on the weighing method after
being dried at 103–105 °C and burnt to ash at 550 °C. Methane
(CH4) production was determined using a gas chromatograph
(Clarus 580 Arnel, PerkinElmer, USA) equipped with a thermal con-
ductivity detector. Volatile fatty acids (VFA) such as acetic acid,
propionic acid and butyric acid were determined by a gas chro-
matograph (Clarus 580GC, PerkinElmer, USA) equipped with a
flame ionization detector. C and N elemental analyses in FW were
determined using the vario MICRO cube (Elementar, Germany).
The abundance of bacteria and archaea were determined by real-
time PCR according to the method described by Zhang et al. [24].
 J. Zhang et al. / Applied Energy 194 (2017) 287–295
Fig. 1. Schematic diagram of three-stage anaerobic digester (TSAD) system.
2.4. DNA extraction and high-throughput 16S rDNA gene
pyrosequencing
The genomic DNA of the sample was extracted using an extrac-
tion kit (MO BIO Laboratories, Inc. Carlsbad, USA) according to the
manufacturer’s instructions. The quality of the extracted DNA was
checked by determining its absorbance at 260 nm and 280 nm.
The diversity of microbial communities was examined by
IIIumina Hiseq 2000 pyrosequencing technology. A set of bacterial
primers 341F (50 -CCTACGGGNGGCWGCAG-30 ) and 805R (50 -GAC
TACHVGGGTATCTAATCC-30 ) was used to amplify the hypervariable
V3–V4 region of bacterial 16S rRNA gene using nested PCR. The
first round archaeal primers was 340F (50 -CCCTAYGGGGYGCAS
CAG-30 ) and 1000 R (50 -GGCCATGCACYWCYTCTC-30 ). The second
round archaeal primers was 349F (50 -GYGCASCAGKCGMGAAW-30
) and 806 R (50 -GGACTACVSGGGTATCTAAT-30 ). After being purified
and quantified, the PCR products of a V3–V4 region of 16S rRNA
gene were determined by pyrosequencing using the IIIumina Hiseq
2000 sequencer (Sangon Biotech Shanghai Co., Ltd., People’s
Republic of China). To obtain the effective sequencing data, raw
289
pyrosequencing results was processed as follows: (1) check the
completeness of the barcodes and the adapter; (2) remove
sequences containing ambiguities (‘‘Ns”); (3) remove sequences
shorter than 200 bps, and (4) remove low - quality sequence i.e.
a sequencing quality value lower than 20 [25]. Subsequently, effec-
tive sequences were clustered into operation taxonomic unit
(OTUs) by a 3% or 5% distance level [26]. Rarefaction curves, Shan-
non diversity index, species richness estimator of Chao1 and
Coverage index were conducted by MOTHUR to identify the species
diversity for each sample. The OTUs defined by a 3% distance level
were classified using the RDP-II classifier at a 50% confidence
threshold.
3. Results and discussion
3.1. Baseline comparison of SL1, SL2, and TSAD
To evaluate the performance of TSAD on enhancing FW treat-
ment, bench scale TSAD and two control reactors (SL1 and SL2)
290 J. Zhang et al. / Applied Energy 194 (2017) 287–295

were operated in parallel at increasing OLR from 1.6 to 10.0 g

VSFW/L. Fig. 2 shows the variations of methane yield, methane pro-
duction rate (MPR), pH and VFA concentration.
At an OLR of 1.6 g VSFW/L, the accumulated CH4 yield in the
three reactors varied between 2.8 and 3.7 L, and pH ranged from
7.2 to 7.4. Due to the low initial OLR, these differences were hardly
significant. The MPR in TSAD was almost the same as that of SL1
and SL2 (Fig. 2B) in the initial 20 days of OLR up to 2.41 g
VSFW/L. With further increase in OLR, the MPR in TSAD increased
monotonically from 0.4 to 1.0 L CH4/d at OLR of 8.02 g VSFW/L.
When the OLR was increased to 8.02 g VSFW/L, the accumulated
methane yield in TSAD increased to 48.1 L and the average pH
was maintained at 7.8. The methane yield was only 36.2 L and
43.4 L in SL1 and SL2, respectively, and the pH had dropped to
5.5 and 7.4, respectively. These results indicate that the methano-
genic activity in TSAD was much higher than that of SL1 and SL2.
From Fig. 2D, the VFA concentration in SL1 started at a low level
initially but increased rapidly to 15,009 mg COD/L at an OLR of
8.02 g VSFW/L, significantly higher than the 1525 mg COD/L and
2623 mg COD/L in TSAD and SL2, respectively. As a key process
parameter in AD, high VFA concentration is typically an indication
of process instability [27]. The sharp decrease of pH in SL1 was
attributed to the accumulation of VFA that happened when the
reactor was fed a high OLR. It has been reported that VFA accumu-
lation which is a possible reason for the decline of MPR in SL1 and
SL2. When the OLR was raised further to 10.0 g VSFW/L, accumu-
lated methane yield in TSAD reached 56.8 L, which was signifi-
cantly higher than 46.0 L in SL2 and 36.8 L in SL1, a 24% and 54%
increase, respectively, by comparison. The pH in TSAD held con-
stant at 7.3 while it decreased to 5.1 in SL1 and 6.0 in SL2, respec-
tively. Along with the decrease of pH, the accumulation of VFA in
SL1 and SL2 reached 35,875 mg COD/L and 30,089 mg COD/L
respectively, 6 times higher than that of TSAD. Most significantly,
at this high OLR, TSAD still continued to perform a high MPR of
1.1 L CH4/d, the MPR in SL1 and SL2 had decreased sharply to
0.07 L CH4/d and 0.3 L CH4/d, respectively.

3.2. Hydrolytic and acidogenic performance in TSAD system

The improved methane yield in TSAD over SL1 and SL2 might be
attributed to the enhanced high-solids hydrolysis and acidogenesis
before the feed was introduced to TSAD. These can be evaluated by
the degree of solubilization and acidification, respectively. The
degree of solubilization can be expressed as the quotient between
soluble COD and total COD of solid sample [28]. The degree of acid-
ification is defined as the ratio of COD-equivalent TVFA to the sol-
uble COD [29]. As shown in Table 1, the degree of solubilization
after high-solids hydrolysis increased from 28 to 42% and the
degree of acidification from 29 to 45%. After acidogenesis, the
degree of solubilization and acidification increased further to 52%
and 66%, respectively. It is apparent that more VFAs were produced
after the acidogenesis stage, which was more favorable for
methane production as compared with other complex soluble sub-
strates like polysaccharides and solid FW. In addition, Moreover,
the conversion of VFA to acetate might have contributed partially
to the enhanced buffering ability of TSAD.
As an important intermediate product from protein degrada-
tion, soluble ammonia concentration increased to 2.33 ± 0.04 and
6.22 ± 0.09 mg/g VS, respectively, after high-solids hydrolysis and
acidogenesis process. The increase of ammonia concentration also
facilitated improve the buffering ability of TSAD according to the
reaction: CxHyCOOH + NH3H2O ? CxHyCOO + NH+4 + H2O [30] Fig. 2. Comparison of (A) accumulated methane yield, (B) methane production rate,
and prevented excess pH decrease. It, therefore, seems plausible (C) pH and (D) TVFA concentration among the three ADs (SL1, SL2 and TSAD).
 J. Zhang et al. / Applied Energy 194 (2017) 287–295
Table 1
Characteristics of feedstocks after hydrolysis and acidogenesis.
c c
Parameter SCOD
(g COD/g VS)
Fresh food waste 0.31 ± 0.02
Food waste after 1st stage high solids hydrolysis 0.47 ± 0.02
Food waste after 2nd stage hydrolysis and acidogenesis 0.58 ± 0.03
a
Degree of solubilization is expressed as the quotient between soluble COD (SCOD) and total COD (TCOD) [21].
b
Degree of acidification is defined as the ratio of COD-equivalent TVFA to the SCOD [22].
c
Data are the averages of the values obtained. Error bars represent standard deviations of statistical analysis.
that the enhanced methane potential in TSAD resulted from the
increased degree of solubilization and acidification, and the effec-
tive buffering in the reactor.
3.3. Overall AD performance and microorganisms quantification
Table 2 summarizes the performance of the three baseline reac-
tors in terms of methane yield and composition, VSFW reduction,
and the composition of the microorganisms after 74 days of oper-
ation. Theoretically, anaerobic methane production was stoichio-
metrically (i.e. maximum) 0.35 L/g CODremoved [31]. Based on
VSFW and COD data presented in Table 1, theoretical conversion
of CH4 can be determined as 0.392 L CH4/g VSFW. It can be seen
from Table 2 that methane yield from TSAD was 0.307 L CH4/g
VSFW after 74 days of operation, which was rather close to the max-
imum theoretical yield. The losses of yield can be attributed to the
conversion of biomass, as well as incomplete reduction of VS in the
feedstock (Table 2). Nevertheless, the methane yield in TSAD is
much higher than that obtained from SL1 and SL2. Furthermore,
the biogas quality, in terms of CH4 content, was also higher in TSAD
(52–67%) compared with <10% in SL1 and <33% in SL2.
Real-time PCR analysis showed that the number of copies of the
archaea in TSAD was 1.48  108 copies/ng-DNA, showing 4 times
and 9 times higher abundance than that of SL2 and SL1, respec-
tively (Table 2). Archaea presents most of the methanogens. There-
fore, higher abundance of archaea would be a good biological
reason for the improved methane yield in TSAD. In comparison,
SL1 presented the highest numbers of copies of bacteria among
the three reactors, while its archaeal abundance was the lowest.
It is precisely this difference in the relative abundance between
bacteria and archaea that is liable for the difference between
hydrolysis, acidogenesis and methanogenesis in the 3 reactors. To
investigate the function of bacteria and archaea in sufficient detail,
microbial community analysis undertaken.
3.4. Analysis of the microbial community
3.4.1. Bacterial community
After pyrosequencing for bacterial community analysis, the
effective sequence numbers of each sample from SL1, SL2, TSAD
and seed sludge were 28,305, 26,053, 27,840 and 38,769 respec-
tively. As shown in Table 3, the observed number of operational
taxonomic units (OTUs) at a 3% distance level were 1189 (SL1),
1596 (SL2), 1645 (TSAD) and 7377 (seed sludge), respectively, indi-
Table 2
VS reduction, 16S rRNA gene copies of bacteria and archaea, methane yields and methane content in the three anaerobic digesters after 74 days of operation.
Digester Biogas
Yield L CH4/g VSFW CH4 content (%)
SL1 0.199 ± 0.015 <10
SL2 0.249 ± 0.008 <33
TSAD 0.307 ± 0.006 52–67
Values are expressed as mean value ± standard deviations.
291
c c a b
TCOD TVFA Ammonia Degree of Degree of
(g COD/g VS) (g COD/g VS) (mg/g VS) solubilization (%) acidification (%)
1.12 ± 0.05 0.09 ± 0.01 0.78 ± 0.02 28 29
1.12 ± 0.05 0.21 ± 0.02 2.33 ± 0.04 42 45
1.12 ± 0.05 0.38 ± 0.03 6.22 ± 0.09 52 66
cating that bacterial community was richer in TSAD compared to
SL1 and SL2. The enhanced biodiversity might be attributed to
the positive effects of high-solids hydrolysis and acidogenesis stage
of TSAD, during which more available substrates e.g. VFAs and
ammonia, were produced for the growth of anaerobic bacteria. It
was reported that high biodiversity contributed to the enhance-
ment of ecological stability on many levels [32], facilitating the
degradation of refractory substance in AD and resulting in high
treatment performance [33,34]. Thus, the enhanced biodiversity
in TSAD might drive the mineralization of complex organic matters
and resist environmental express e.g. high OLR shock. However,
the richness of bacterial community in these three reactors
decreased sharply as compared with seed sludge, suggesting that
only partial dominant species could be enriched while most of
the bacteria were eliminated during the reactor operation. Alpha
diversity analysis in terms of Shannon index and ACE index
showed higher heterogeneity in TSAD. From the hierarchical clus-
ter analysis in Fig. 3, the bacterial community in SL1 and SL2 was
classified to one cluster. This cluster was separated from the TSAD
cluster and seed sludge, suggesting that high-solids hydrolysis and
acidogenesis process in TSAD significantly changed the bacterial
community structure. The heterogeneity and difference might
serve the bacterial community of TSAD both for high biodegrada-
tion capacity and resistibility against complex organic wastes
and high OLR. In comparison, low biodiversity in SL1 and SL2
caused low treatment efficiency and methane yield (Fig. 2). To fur-
ther investigate the composition of the bacterial community, bac-
terial taxonomy analysis was performed.
As shown in Fig. 4, the species taxonomy of the OTU of SL1, SL2,
TSAD and seed sludge were classified at the genus level, relative
abundance higher than 1% was defined as the dominant species.
Fourteen dominant genus were identified in TSAD except uncul-
tured bacteria, in which 8 dominant genus i.e. Ruminococcus
(6.74%), Anaerosphaera (1.44%), Sporanaerobacter (1.13%),
Bacteroides (4.85%), Syntrophomonas (2.44%), Levilinea (15.27%),
Aminomonas (1.13%), Proteiniphilum (9.93%) showed the highest
relative abundance among the three ADs. Sporanaerobacter and
Proteiniphilum belong to acetogens which were capable of
acetogenesis and that accelerated the transfer of VFAs to acetate
e.g. CH3CH2COOH + 2H2O ? CH3COOH + CO2 + 3H2 [19,35]. As is
known, only acetic acid and H2/CO2 could be used by methanogens
for methane production. Therefore, the enrichment of bacterial
species Proteiniphilum and Sporanaerobacter in TSAD might have
contributed to enhanced acetogenesis and further improved
VS reduction (%) 16S rRNA gene copies/g TS
Bacteria Archaea
30.1 ± 3.2 1.22  109 1.68  107
44.2 ± 2.9 9.94  108 3.44  107
83.5 ± 1.3 8.94  108 1.48  108
292 J. Zhang et al. / Applied Energy 194 (2017) 287–295

Table 3
Biodiversity estimation of 16S rRNA gene libraries from the pyrosequencing analysis.

Parameters Effective reads Observed OTUs Shannon ACE Coverage

SL1 28,305 1189 4.17 3141 0.98
SL2 26,053 1596 4.61 4905 0.97
TSAD 27,840 1645 4.68 4991 0.97
Seed sludge 38,769 7377 7.23 11,555 0.92

Seed sludge TSAD SL1 SL2

Fig. 3. Hierarchical cluster analysis of bacterial communities in the sludge samples of SL1, SL2, TSAD and seed sludge. The OTUs were instructed at phylum level. Sample
communities were clustered according to the method of complete linkage. The color intensity of scale indicates relative abundance of each OTU read. Relative abundance was
defined as the numbers of sequences affiliated with that OTU divided by the total number of sequences per sample.

methanogenesis. Both Anaerosphaera and Aminomonas belong to 3.4.2. Archaeal community

amino-acid-utilizing bacteria that degraded the amino-acids and
produced acetate, propionate and ammonia, etc. as end-products
[36]. Ammonia ensured sufficient buffer capacity of AD and kept
a neutral pH thus increasing the stability of AD [37], a possible
main reason for the high performance and methane yields in TSAD.
Syntrophomonas, as an obligately anaerobic and syntrophic bacte-
ria, has the ability to oxidize saturated fatty acids, which is usually
in syntrophic association with H2-using methanogen and enhances
the enrichment of H2-using methanogen. Bacteroides is a genus of
obligately anaerobic, playing a fundamental role in the processing
of complex molecules to simpler compounds. Ruminococcus is a
genus of bacteria in the class Clostridia, which belonged to fermen-
tative bacteria with acetic and propionic acid as its major acidifica-
tion products. Additionally, the relative abundance of genus
Saccharofermentans, Petrimonas, Aminobacterium and Eubacterium
in TSAD and SL2 were much higher than that in SL1. However, it
was observed that many genera were washed out from the seed
sludge after long-term operation of AD e.g. Sphingomonas,
Acinetobacter and Pseudomonas, etc.
A total number of 33,257 (SL1), 37,835 (SL2), 32,751 (TSAD) and
20,067 (seed sludge) effective reads were obtained by the pyrose-
quencing. Fig. 5 shows the methanogenic microbial composition at
the genus level. Compared to the bacterial communities (Fig. 4),
the richness of archaeal communities in these three reactors was
very low. Only four predominant archaeal genus i.e. Methanosar-
cina, Methanospirillum, Methanospirillum and Methanosaeta were
found to be dominant in these three reactors, which accounted
for more than 99.2% of the total archaeal population. The relative
abundance of Methanosarcina in TSAD was 31.2% of the total
archaeal population, which was significantly higher than the
22.7% in SL1 and 20.3% in SL2. It is known that Methanosarcina is
a multifunctional methanogen that can produce methane via three
different metabolic pathway using H2/CO2, acetate and methylated
one-carbon compounds [38]. The intermediate products from
anaerobic digestion of complex organic matter are diverse that
the utilization of all these products are limited because most
methanogens can only use sole substrates such as acetate or
H2/CO2. Therefore, the enrichment of Methanosarcina might have
 J. Zhang et al. / Applied Energy 194 (2017) 287–295 293

Fig. 4. Taxonomic classification of the dominant phylogenetic groups from SL1, SL2, TSAD and Seed sludge at the genus levels.

Fig. 5. Taxonomic classification of the dominant archaeal groups from SL1, SL2, TSAD and Seed sludge at the genus levels.

enhanced methane production by using various substrates from ment with the results of methane yields (Fig. 2). The sum of the rel-
high-solids hydrolysis and acidogenesis of TSAD that was in agree- ative abundance of Methanobacterium and Methanospirillum in SL1,
294 J. Zhang et al. / Applied Energy 194 (2017) 287–295
Fig. 6. (A) A 20 L small-scale three-stage anaerobic digester; (B) Methane yield and pH.
SL2, TSAD and seed sludge were 14%, 18.2%, 30.9% and 9.9%, respec-
tively, both of which belonged to hydrogen-utilizing methanogen.
In comparison, the abundance of acetate-utilizing methanogen i.e.
Methanosaeta in SL1 and SL2 were much higher.
3.5. Preliminary test of a small-scale TSAD
All of the baseline experiments compared the performance
between TSAD and the two control reactors. TSAD presented the
highest performance for food waste treatment and methane pro-
duction. Generally, the ideal methane yield from FW in lab-scale
test ranges between 348 ml/g VSFW and 455 mg/g VSFW [14,30].
However, the practical methane yield in some small-scale or
large-scale tests is even less than 200 ml/g VSFW due to the com-
plexity of the practical operating process and feedstocks and
imperfect conditions [39–41]. To shorten the timespan between
lab-scale test and application, a small-scale 20 L TSAD (Fig. 6A)
was fabricated and tested. The performance of the TSAD was tested
for food waste treatment. The operational conditions of this small-
scale TSAD were the same as the 1 L bench scale AD. From Fig. 6B,
the operation of the small-scale TSAD was rather stable at the OLR
of 2 g VSFW/L. After 30 days of operation, the methane yield
reached 0.34 L/g VSFW d, and pH was kept at 7. The VS reduction
of this small-scale TSAD reached around 81%, consistent with the
report of Kiran et al. [41] who achieved 80.4% VS removal for AD
of FW treatment. Compared to the ideal methane yield [14,30],
there is room for improvement. In Singapore, the amount of FW
generated in 2015 was about 785,500 tonnes [42]. It can be
assumed that the VS content of the FW is 17–32% [23]. In terms
of real applications therefore, if all of the FW generated can be
digested by TSAD, the methane yield can reach 4.5  107–
8.55  107 m3/year. Bench-scale and prototype-scale tests have
been initiated to evaluate the feasibility of accomplishing this.
These tests haven been conducted with the aim to optimize the
reactor configuration and construction, operational parameters
including OLR, pH control and temperature, and inoculum
selection.
4. Conclusions
A three-stage anaerobic digester has been fabricated for
enhanced FW treatment and methane production. The functional-
ized partitioning in TSAD significantly enhanced the hydrolysis and
acidogenesis efficiency of solid FW, resulting in a 24–54% increase
in methane production. Due to the optimized inner structure, a
higher VS removal of 83.5 ± 1.3% was achieved in TSAD. Compared
to the traditional anaerobic digesters, TSAD had higher compact
structure and buffering ability even exposed to high OLR. This work
provides an alternative approach to enhance organic solid waste
treatment and biochemical methane potential.
Acknowledgments
This research/project is supported by the National Research
Foundation, Prime Minister’s Office, Singapore under its Campus
for Research Excellence and Technological Enterprise (CREATE)
programme.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.apenergy.2016.
10.116.
References
[1] Zhang K, Pei Z, Wang D. Organic solvent pretreatment of lignocellulosic
biomass for biofuels and biochemicals: a review. Bioresour Technol
2016;199:21–33.
[2] Li YY, Jin YY, Li JH, Li HL, Yu ZX. Effects of pungency degree on mesophilic
anaerobic digestion of kitchen waste. Appl Energy 2016;181:171–8.
[3] Zhang C, Su H, Baeyens J, Tan T. Reviewing the anaerobic digestion of food
waste for biogas production. Renew Sustain Energy Rev 2014;38:383–92.
[4] Mao C, Feng Y, Wang X, Ren G. Review on research achievements of biogas
from anaerobic digestion. Renew Sustain Energy Rev 2015;45:540–55.
[5] Borjesson P, Berglund M. Environmental systems analysis of biogas systems –
Part I: fuel-cycle emissions. Biomass Bioenergy 2006;30:469–85.
[6] Poeschl M, Ward S, Owende P. Evaluation of energy efficiency of various biogas
production and utilization pathways. Appl Energy 2010;87:3305–21.
[7] Sansartier D, MacLean HL, Saville B. Electricity production from anaerobic
digestion of household organic waste in Ontario: techno-economic and GHG
emission analyses. Environ Sci Technol 2012;46:1233–42.
[8] Raposo F, De la Rubia MA, Fernández-Cegrí V, Borja R. Anaerobic digestion of
solid organic substrates in batch mode: an overview relating to methane yields
and experimental procedures. Renew Sustain Energy Rev 2012;16:861–77.
[9] Kothari R, Pandey AK, Kumar S, Tyagi VV, Tyagi SK. Different aspects of dry
anaerobic digestion for bio-energy: an overview. Renew Sustain Energy Rev
2014;39:174–95.
[10] Fernández J, Pérez M, Romero LI. Effect of substrate concentration on dry
mesophilic anaerobic digestion of organic fraction of municipal solid waste
(OFMSW). Bioresour Technol 2008;99:6075–80.
[11] Zhen GY, Lu XQ, Kobayashi T, Li YY, Xu KQ, Zhao YC. Mesophilic anaerobic co-
digestion of waste activated sludge and Egeria densa: performance assessment
and kinetic analysis. Appl Energy 2015;148:78–86.
[12] Zhang WQ, Wei QY, Wu SB, Qi DD, Li W, Zuo Z, Dong RJ. Batch anaerobic co-
digestion of pig manure with dewatered sludge under mesophilic conditions.
Appl Energy 2014;128:175–83.
[13] Li Y, Park SY, Zhu J. Solid-state anaerobic digestion for methane production
from organic waste. Renew Sustain Energy Rev 2011;15:821–6.
[14] Angelidaki I, Ahring BK. Effects of free long-chain fatty-acids on thermophilic
anaerobic-digestion. Appl Microbiol Biotechnol 1992;37:808–12.
 J. Zhang et al. / Applied Energy 194 (2017) 287–295
[15] Buffiere P, Loisel D, Bernet N, Delgenes JP. Towards new indicators for the
prediction of solid waste anaerobic digestion properties. Water Sci Technol
2006;53:233–41.
[16] Schievano A, D’Imporzano G, Malagutti L, Fragali E, Ruboni G, Adani F.
Evaluating inhibition conditions in high-solids anaerobic digestion of organic
fraction of municipal solid waste. Bioresour Technol 2010;101:5728–32.
[17] Yang L, Xu F, Ge X, Li Y. Challenges and strategies for solid-state anaerobic
digestion of lignocellulosic biomass. Renew Sustain Energy Rev
2015;44:824–34.
[18] Kim J, Park C, Kim TH, Lee M, Kim S, Kim SW, Lee J. Effects of various
pretreatments for enhanced anaerobic digestion with waste activated sludge. J
Biosci Bioeng 2003;95:271–5.
[19] Lee DH, Behera SK, Kim JW, Park HS. Methane production potential of leachate
generated from Korean food waste recycling facilities: a lab-scale study. Waste
Manage 2009;29:876–82.
[20] Harris PW, McCabe BK. Review of pre-treatments used in anaerobic digestion
and their potential application in high-fat cattle slaughterhouse wastewater.
Appl Energy 2015;155:560–75.
[21] Ariunbaatar J, Panico A, Esposito G, Pirozzi F, Lens PNL. Pretreatment methods
to enhance anaerobic digestion of organic solid waste. Appl Energy
2014;123:143–56.
[22] Ge XM, Xu FQ, Li YB. Solid-state anaerobic digestion of lignocellulosic biomass:
recent progress and perspectives. Bioresour Technol 2016;205:239–49.
[23] Dai X, Duan N, Dong B, Dai L. High-solids anaerobic co-digestion of sewage
sludge and food waste in comparison with mono digestions: stability and
performance. Waste Manage 2013;33:308–16.
[24] Zhang J, Zhang Y, Quan X. Bio-electrochemical enhancement of anaerobic
reduction of nitrobenzene and its effects on microbial community. Biochem
Eng J 2015;94:85–91.
[25] Li S, Chou HH. Lucy 2: an interactive DNA sequence quality trimming and
vector removal tool. Bioinformatics 2004;20:2865–6.
[26] Cole JR, Chai B, Farris RJ, Wang Q, Kulam SA, McGarrell DM, Garrity GM, Tiedje
JM. The Ribosomal Database Project (RDP-II): sequences and tools for high-
throughput rRNA analysis. Nucleic Acids Res 2005;33:D294–6.
[27] Ahring BK, Sandberg M, Angelidaki I. Volatile fatty-acids as indicators of
process imbalance is anaerobic digesters. Appl Microbiol Biotechnol
1995;43:559–65.
[28] Demirel B, Yenigun O. Anaerobic acidogenesis of dairy wastewater: the effects
of variations in hydraulic retention time with no pH control. J Chem Technol
Biotechnol 2004;79:755–60.
[29] De La Rubia MA, Raposo F, Bincon B, Borja R. Evaluation of the hydrolytic-
acidogenic step of a two-stage mesophilic anaerobic digestion process of
sunflower oil cake. Bioresour Technol 2009;100:4133–8.
295
[30] Zhang R, El-Mashad HM, Hartman K, Wang F, Liu G, Choate C, Gamble P.
Characterization of food waste as feedstock for anaerobic digestion. Bioresour
Technol 2007;98:929–35.
[31] Toprak H. Temperature and organic loading dependency of methane and
carbon-dioxide emission rates of a full-scale anaerobic waste stabilization
pond. Water Res 1995;29:1111–9.
[32] Wrighton KC, Agbo P, Warnecke F, Weber KA, Brodie EL, DeSantis TZ,
Hugenholtz P, Andersen GL, Coates JD. A novel ecological role of the
Firmicutes identified in thermophilic microbial fuel cells. ISME J
2008;2:1146–56.
[33] Liu WZ, Wang AJ, Cheng SA, Logan BE, Yu H, Deng Y, Nostrand JDV, Wu LY, He
ZL, Zhou JZ. Geochip-based functional gene analysis of anodophilic
communities in microbial electrolysis cells under different operational
modes. Environ Sci Technol 2010;44:7729–35.
[34] Lu L, Xing DF, Ren NQ. Pyrosequencing reveals highly diverse microbial
communities in microbial electrolysis cells involved in enhanced H2
production from waste activated sludge. Water Res 2012;46:2425–34.
[35] Chen SY, Dong XZ. Proteiniphilum acetatigenes gen. nov., sp nov., from a UASB
reactor treating brewery wastewater. Int J Syst Evol Microbiol
2005;55:2257–61.
[36] González-Fernández C, García-Encina PA. Impact of substrate to inoculum
ratio in anaerobic digestion of swine slurry. Biomass Bioenergy
2009;33:1065–9.
[37] Rajagopal R, Massé DI, Singh G. A critical review on inhibition of anaerobic
digestion process by excess ammonia. Bioresour Technol 2013;143:632–41.
[38] Galagan JE, Nusbaum C, Roy A, Endrizzi MG, Macdonald P, FitzHugh W, Calvo S,
Engels R, et al. The genome of M-acetivorans reveals extensive metabolic and
physiological diversity. Genome Res 2002;12:532–42.
[39] Ahamed A, Chen CL, Rajagopal R, Wu D, Mao Y, Ho IJR, Lim JW, Wang JY. Multi-
phased anaerobic baffled reactor treating food waste. Bioresour Technol
2015;182:239–44.
[40] Chen GY, Liu G, Yan BB, Shan R, Wang JN, Li T. Experimental study of co-
digestion of food waste and tall fescue for bio-gas production. Renewable
Energy 2016;88:273–9.
[41] Kiran EUK, Trzcinski AP, Liu Y. Enhancing the hydrolysis and methane
production potential of mixed food waste by an effective enzymatic
pretreatment. Bioresour Technol 2015;183:47–52.
[42] Waste and Recycling Statistics for 2015, <www.nea.gov.sg/energy-waste/
waste-management/waste-statistics-and-overall-recycling>, National
Environment Agency of Singapore.