45 © IWA Publishing 2015 Water Science & Technology | 72.

1 | 2015

Influence of denitrification reactor retention time

distribution (RTD) on dissolved oxygen control and
nitrogen removal efficiency
Massimo Raboni, Renato Gavasci and Paolo Viotti

ABSTRACT
Massimo Raboni
Low concentrations of dissolved oxygen (DO) are usually found in biological anoxic pre-denitrification
School of Industrial Engineering,
reactors, causing a reduction in nitrogen removal efficiency. Therefore, the reduction of DO in such University LIUC – Cattaneo,
Corso Matteotti, 22, I-21053 Castellanza, Varese,
reactors is fundamental for achieving good nutrient removal. The article shows the results of an Italy

experimental study carried out to evaluate the effect of the anoxic reactor hydrodynamic model on Renato Gavasci
Department of Civil Engineering and Computer
both residual DO concentration and nitrogen removal efficiency. In particular, two hydrodynamic Science Engineering,
University of Rome ‘Tor Vergata’,
models were considered: the single completely mixed reactor and a series of four reactors that
Via Politecnico 1, I-00133 Roma,
resemble plug-flow behaviour. The latter prove to be more effective in oxygen consumption, allowing Italy

a lower residual DO concentration than the former. The series of reactors also achieves better Paolo Viotti (corresponding author)
Department of Civil and Environmental
specific denitrification rates and higher denitrification efficiency. Moreover, the denitrification food to Engineering,
University of Roma La Sapienza,
microrganism (F:M) ratio (F:MDEN) demonstrates a relevant synergic action in both controlling residual Via Eudossiana 18, I-00184 Rome,
Italy
DO and improving the denitrification performance.
E-mail: paolo.viotti@uniroma1.it
Key words | completely mixed reactor, denitrification rate, hydrodynamic model, residual
dissolved oxygen

INTRODUCTION

Denitrification is the biological process where nitrite (NO3–N)
is converted into gaseous nitrogen (N2) under anoxic con-
ditions. Considering the pre-denitrification process, the reactor
design is usually based on the denitrification rate (rDEN),
assuming a zero-order kinetics with respect to both NO3–N
and organic substrate, and considering the effect of temperature
W
(T ). Typical values of rDEN at 20 C are in the range 2.9–3.0
gNO3–N kgMLVSS
1 h
1 (MLVSS ¼ mixed-liquor volatile
suspended solids) (Barnard ; Sutton et al. ; US-EPA
; De Fraja Frangipane et al. ; Ekama ; Vismara
; Ekama & Wentzel ; Ekama ; Collivignarelli &
Bertanza ; Torretta et al. ).
Dissolved oxygen (DO) is an inhibiting factor for denitri-
fication, and concentrations in the range 0.2–0.4 mgO2 L
1
The possible inhibitory effects of DO on the denitrifica-
tion kinetics were postulated by US-EPA () and were
highlighted by other studies (Badstreet & Johnson ;
Oh & Silverstein ; Plòsz et al. ). The presence of
0.2 mgO2 L
1 lowers rDEN up to 40% of the maximum
values measured without DO (Dawson & Murphy ;
Tchobanoglous et al. ). To take account of such an
effect, inhibition factors were introduced in various rDEN
models, as described elsewhere (US-EPA , ;
Luciano et al. ; Fajardo et al. ).
For practical calculation of the denitrification reactor
(DEN) volume, a semi-empirical relationship, which corre-
lates the specific denitrification rate at T ¼ 20 C
W
(SDNR20 C) to the sludge loading referred to DEN, has
W

significantly reduce rDEN. DO is even higher in small waste- been proposed (Burdick et al. ; US-EPA ; Tchobano-
water treatment plants that are characterized by strong glous et al. ):
changes of sewage flow and quality (Raboni et al. ).
DO presence within anoxic reactors depends on two oppos- SDNR20 C ¼ 0:03  F:MDEN þ 0:029
W
(1)
ing factors: (i) DO content into the inflow (sewage and
mixed-liquor recycle); and (ii) oxygen consumption deter- where SDNR ¼ (Q · ΔNO3–N)/(VDEN · X); Q is the sewage
mined by the heterotrophic bacteria. flow rate (m3 d
1); ΔNO3–N is the NO3–N removed into
doi: 10.2166/wst.2015.188
 46 M. Raboni et al. | Influence of retention time in denitrification process
DEN (gNO3–N m
3); VDEN is the DEN volume (m3); X is
the biomass concentration in DEN (gMLVSS m
3); F:
MDEN is the sludge loading referred to DEN (gBOD5 applied
to gMLVSS
1 d
1). Values of SDNR observed in full-scale
plants range from 0.04 to 0.42 gNO3–N gMLVSS
1 d
1
(Burdick et al. ; Henze ; Tchobanoglous et al.
), while US-EPA () reports a more restricted range
(0.05–0.15 gNO3–N gMLVSS
1 d
1 at 20 C).
W
As the temperature strongly affects the denitrification
kinetics, the well-known correlation has been proposed:
SDNRT ¼ SDNR20 W
C  θ ðT
20Þ
where SDNRT is the SDNR at a generic temperature T; θ ¼
1.026 for Tchobanoglous et al. () and θ ¼ 1.07 for US-
EPA ().
More recently, Raboni et al. () highlighted the strong
dependence of pre-denitrification SDNR on both DO and
food to microrganism (F:M) ratio. They proposed the follow-
ing equation:
SDNR20 C W


K0 O
¼ 0:0864 
K0 O þ DO0  ð1
[(K  B0 )=(F :MDEN )]Þ
þ 0:05 F:MDEN  ηBOD


1
 (3)
[0:2=(DO0  (1
[(K  B0 )=(F : MDEN )]))] þ 1
where K0 O ¼ 0.18 mgO2 L
1 is the DO inhibition constant;
DO0 and B0 are the DO and biochemical oxygen demand
(BOD5) concentrations of the whole flow rate entering the
DEN (mg L
1); K ¼ 0.11–0.18 is a constant; ηBOD is the
BOD removal efficiency, which depends on the F:MDEN
(ηBOD ¼ 0.90 for F:MDEN ¼ 0.4 kgBOD5 kgMLVSS
1 d
1;
ηBOD ¼ 0.95 for F:MDEN ¼ 0.2 kgBOD5 kgMLVSS
1 d
1).
They also proposed a more general equation based on the
experimental correlation between SDNR20 C, DO and
BOD5 detected in the total flow rate entering the DEN.
Having ascertained the considerable dependence of the
denitrification rate on DO concentration, the paper verifies
the effect of the DEN hydrodynamics on the residual con-
centration of DO and, consequently, on the process
performance. A research experiment with two pilot
plants, which differ from the hydrodynamic model of the
anoxic reactor, has been developed. In the first plant, a
series of four completely mixed reactors was implemented
in order to simulate a hydrodynamic behaviour approach-
ing the ‘plug-flow’, while in the second case, a single
completely mixed reactor was chosen. Both solutions
Water Science & Technology | 72.1 | 2015
represent typical situations found in full-scale plants. The
input of DO in the completely mixed reactor (mainly
related to the mixed-liquor recycle) is immediately dis-
persed throughout the volume, and its consumption due
to the heterotrophic bacteria is low. In contrast, the
system of the four reactors in series allows a descending
DO concentration from very high values in the first reactor
to very small values in the last reactor. Such solutions are
beneficial to denitrification efficiency because they
increases DO consumption.
(2)
METHODS
Pilot plant description
The research was based on two parallel activated sludge pilot
plants (Figure 1), consisting of a biological pre-denitrification
stage (DEN; volume: 10 m3; liquid height: 1.8 m), an
oxidation-nitrification stage (OX-NIT; volume: 20 m3; liquid
height: 1.8 m; aeration system: micro-bubble), and a final
sedimentation (SED; volume: 6 m3; diameter: 2 m). The
plants differ in the DEN configuration:
• Pilot plant 1 (Figure 1(a)) has a series of four identical
completely mixed reactors (volume of a single reactor:
2.5 m3). Mixing is achieved by four slow vertical-axis
mixers (power input: 11 W m
3).
• Pilot plant 2 (Figure 1(b)) has a single completely mixed
reactor. Mixing is achieved by one slow vertical-axis
mixer (power input: 16 W m
3).
DEN water surfaces were covered with a 1 inch layer of
floating plastic balls in order to reduce the oxygen mass
transfer between atmosphere and water.
The instrumentation with continuous sampling for plant
W
control consists of (Figure 2): 16 DO fixed probes (accuracy:
0.01 mg L
1; automatic calibration and temperature com-
pensation); six pH fixed probes (accuracy: 0.05); four
temperature fixed probes (accuracy: 0.05 C); and four mag-
W
netic flow-meters (accuracy: 0.5% of the flow rate).
The pilot plants were fed by pre-treated (screening and
aerated grit chamber) sewage from a 50,000 inhabitant
town located in Northern Italy.
Pilot plant operating conditions and testing methods
The experimentation was carried out using the following
steps.
 47 M. Raboni et al. | Influence of retention time in denitrification process Water Science & Technology | 72.1 | 2015

Figure 1 | Layout of the two parallel pilot plants: pilot plant 1, having the denitrification stage characterized by a series of four reactors (a); pilot plant 2, having the denitrification stage
characterized by a single completely mixed reactor (b).

Figure 2 | Scheme and instrumentation of the two pilot plants.

• Preliminary verification of the DEN hydrodynamic be-
haviour through reconstruction of the retention time
distribution (RTD) curves (Tchobanoglous et al. ).
A tracer solution of lithium chloride was used with a
step dosage of 5 mgLiþ L
1. For the tests, tap water was
fed to the plants.
• Simultaneous conduction of the two pilot plants with the
same operating conditions was used in order to check:
○ the sewage quality and the overall treatment efficiency
with respect to BOD5, chemical oxygen demand
(COD), total nitrogen (TN) and suspended solids (SS);
 48 M. Raboni et al. | Influence of retention time in denitrification process Water Science & Technology | 72.1 | 2015

○ the DO concentration in the two types of DEN, and its • BOD5, COD, total Kjeldahl nitrogen (TKN), NO3–N
influence on both SDNR and plant denitrification effi- and SS in the pilot plant effluents (daily average
ciency (ηDEN, %), taking into account the role of samplings);
F:MDEN: • MLVSS and mixed-liquor suspended solids (MLSS) in
DEN and OX-NIT (daily manual sampling);
• temperature, DO and pH at the sampling points shown in
TNps
TNeff Figure 2, were measured.
ηDEN ¼  100 (4)
TNps
Sampling and analysis were carried out in compliance
where the subscripts ps and eff refer to the pre-treated with official standard methods of Italian legislation.
sewage and the effluents, respectively.

The two pilot plants ran for a continuous period of

180 d, providing operating controls and analysis. In this RESULTS AND DISCUSSION
period, the F:MDEN was set to:
Raw sewage and treated effluent mean quality
• 0.2 kgBOD5 d
1 kgMLVSS
1 for the first 60 d;
• 0.3 kgBOD5 d
1 kgMLVSS
1 for the subsequent 60 d; Table 1 shows the raw sewage and the treated effluent qual-
• 0.4 kg BOD5 d
1 kgMLVSS
1, for the final 60 d. ity of the two pilot plants during the first 60 d (F:MDEN ¼
0.2 kgBOD5 d
1 kgMLVSS
1). During the experiments, the
The DO in OX-NIT was kept at 2.5 mgO2 L
1 on aver- W
average temperature in DEN was 16.1 C (range: 15.6–
age in both the pilot plants, while the MLVSS W
16.7 C).
concentration was maintained at 2.0 mg L
1. The operating
Data indicate a ‘low strength’ sewage. The average effi-
conditions were as follows:
ciency of the two pilot plants is quite similar (about: 70%
• Average sewage flow rate Q ¼ 2 m3 h
1; for COD; 90% for BOD5; 87% for SS) except for TN,
• Mixed-liquor recycle flow rate QML ¼ 3Q; which shows some significant difference (78.6% for Plant
• Sludge recycle flow rate, q ¼ Q. 1; 71.7% for Plant 2).

The monitored analytical parameters during the exper-

RTD tests
iments were as follows:

• BOD5, COD, TN and SS in the pre-treated sewage (daily
average samplings);
• TN and NO3–N in the total flow rate (sum of Q, QML and
q) both ingoing and outgoing the DEN (daily average
samplings);
Figure 3 shows the results of the tests carried out to verify
the hydrodynamic behaviour of the two DENs.
The anoxic stage of Plant 1 proves to effectively behave
as a series of four completely mixed reactors; in fact, the
experimental curve overlaps almost perfectly (except in the

Table 1 | Quality of the raw sewage and the treated effluent in both plants: mean (m) and standard deviation (sd) of daily samples (number of samples: 21 for sewage; 42 for the treated
effluent)

Treated effluent

Raw sewage Pilot plant 1 Pilot plant 2

Parameter Unit of measurement m sd m sd m sd

1
COD mg L 288.0 59.9 85.5 15.3 88.3 16.1

1
BOD5 mg L 127.9 40.0 12.1 2.5 12.6 2.6
SS mg L
1 153.0 45.0 20.5 4.5 19.7 4.8

1 a a a
TN mg L 27.6 5.1 5.9 2.0 7.8 2.1a
a
All NO3–N because TKN in the effluent was always less than 0.5 mg L
1.
 49 M. Raboni et al. | Influence of retention time in denitrification process Water Science & Technology | 72.1 | 2015

Figure 3 | RTD curves of the two DENs: comparison between experimental and theoretical results.

first 50 min, maybe due to a by-pass effect). The theoretical

curve represented by the formula:

2 3
t N
1
i
6
N X Ni t 7
C ¼ Co  41
e tr  5 (5)
i¼0
i! tr

where C is the tracer concentration in the last reactor efflu-

ent (mgLiþ L
1); Co is the tracer concentration in the first
reactor influent (mgLiþ L
1); N is the number of reactors
in series (N ¼ 4 in the specific case); t is the time (h); tr (h)
is the theoretical hydraulic retention time of the whole
DEN volume (VDEN, m3):

VDEN Figure 4 | DO concentration in the DENs as a function of the sludge loading (F:MDEN). Pilot
tr ¼ (6) plant 1 data are represented for each reactor. The continuous lines and the
Q þ QML þ q shaded areas represent the mean and the 95% confidence interval,
respectively.

Although not perfectly so, the experimental RTD curve

relating to Plant 2 shows that DEN 2 has a behaviour very which is rapidly consumed: DO is almost halved in the
similar to a completely mixed reactor, which is described second stage and it is below 0.1 mgO2 L
1 in the last reactor.
by the formula: The rapid initial drop is caused by the high concentration of
both DO and BOD5, which influence the DO consumption
 
rate as described by the following kinetic equation:
C ¼ Co  1
e
(t=tr ) (7)

rDO ¼ a  rs þ b  X



Pilot plants S DO
¼aKX  þbX (8)
KS þ S KDO þ DO
Influence of the hydrodynamic models on DO
where rDO is the oxygen consumption rate (mgO2 L
1 h
1); rS
Figure 4 shows the DO concentration in the two DENs as a is the substrate removal rate (mg L
1 h
1); S is the substrate
function of the denitrifications sludge loading (F:MDEN). concentration (mg L
1); KS and KDO are the half velocity
Both hydrodynamic models show that the increase of constants referred to substrate and the DO, respectively
sludge loading allows a reduction of the DO concentration. (mg L
1); X is the VSS concentration in the mixed liquor
Considering the configuration with the series of reactors (mg L
1); a, b and K are kinetic constants (dimensionless).
(Plant 1), the first stage has a high DO concentration The average DO concentration in the single completely
(0.61–0.73 mgO2 L
1, depending on the sludge loading) mixed reactor (Pilot plant 2) is less than halved
 50 M. Raboni et al. | Influence of retention time in denitrification process Water Science & Technology | 72.1 | 2015

(0.2–0.3 mgO2 L
1) compared to the initial concentration
found in the four reactors in series and about three times
higher than its final concentration.
Such results demonstrate the good performance in DO
removal of the series of reactors compared to a single reactor.
Influence of the RTD models on both SDNR and
denitrification efficiency
Figure 5 shows the SDNR of the two DEN hydrodynamic
models as a function of the sludge loading in denitrification
(F:MDEN).
The denitrification sludge loading significantly influ-
ences the SDNR: doubling the F:MDEN (from 0.2 to
0.4 kgBOD5 d
1 kgMLVSS
1), the average SDNR increase
is about 54% in both reactor models. In addition, the
hydraulic model affects the denitrification rate: the total
SDNR of the series of four reactors (Plant 1) is 16.6% and
25% higher than the values measured with the single com-
pletely mixed reactor (Plant 2) at F:MDEN equal to 0.2 and
0.4 kgBOD5 d
1 kgMLVSS
1, respectively.
Figure 6 shows the denitrification performance of both
pilot plants during the 180 operational days, expressed as
the overall plant denitrification efficiency.
Data confirm the results of SDNR concerning the influ-
ence of the hydrodynamic model (Figure 5) on nitrogen
removal. The denitrification efficiency with the series of reac-
tors results are 6.7%, 9.7% and 12%, which are higher than for
a single completely mixed reactor for F:MDEN equal to 0.2
kgBOD5 d
1 kgMLVSS
1, 0.3 kgBOD5 d
1 kgMLVSS
1
and 0.4 kgBOD5 d
1 kgMLVSS
1, respectively.

CONCLUSIONS

Dissolved oxygen in the biological DENs represents a serious

limiting factor for the kinetics of the dissimilative reaction
and, consequently, for the process efficiency. Nevertheless,
small concentrations of DO are constantly present in biologi-
cal pre-DENs, where they potentially cause adverse effects.
Figure 5 | SDNR as a function of the sludge loading in denitrification (F:MDEN) for the two
The experimental results obtained with a series of four
W
reactor models. Average temperature: 16.1 C (range: 15.6–16.7 C).
W

reactors proved a capacity of oxygen consumption greater

Figure 6 | Denitrification efficiency (ηDEN) of the two pilot plants at three sludge loadings (F:MDEN).
 51 M. Raboni et al. | Influence of retention time in denitrification process
than a single completely mixed reactor (residual DO lower
than 0.1 mgO2 L
1 compared to 0.18–0.30 mgO2 L
1). Such
results are also positively influenced by high denitrification
sludge loadings (F:MDEN). Tests carried out continuously on
the two parallel pilot plants for 6 months evidenced that the
SDNR raises both using four reactors in series instead of a
single completely mixed reactor and increasing F:MDEN.
Such an improvement amounts to þ16.6% and þ25% (in
favour of the series of four reactors) with F:MDEN equal to
0.2 kgBOD5 d
1 kgMLVSS
1 and 0.4 kgBOD5 d
1

1
kgMLVSS , respectively. The tests demonstrated similar
results considering denitrification removal efficiency (up to
þ12% with F:MDEN ¼ 0.4 kgBOD5 d
1 kgMLVSS
1).
In conclusion, the experience highlights the importance
of the hydrodynamic model of the anoxic reactor in condition-
ing the residual DO concentration and, consequently, the
denitrification performance. For the same purpose, the right
choice of the denitrification F:M ratio is of great importance.
REFERENCES
Badstreet, K. A. & Johnson, G. R.  Study of critical
operational parameters for nitrogen reduction at municipal
wastewater treatment plant. In: Proc. of Water Environment
Federation 67th Annual Conference & Exposition, 15–19th
October 1994, Chicago (IL), USA.
Barnard, J. L.  Biological denitrification. Water Pollution
Control 74, 178–186.
Burdick, C. R., Refling, D. R. & Stensel, H. D.  Advanced
biological treatment to achieve nutrient removal. Journal
Water Pollution Control Federation 54 (7), 1078–1086.
Collivignarelli, C. & Bertanza, G.  Ingegneria Sanitaria-
Ambientale (Sanitary-Environmental Engineering). 1st edn.
Città Studi, Torino, Italy.
Dawson, R. N. & Murphy, K. L.  The temperature dependency
of biological denitrification. Water Resource 6 (1), 71–83.
De Fraja Frangipane, E., Cossu, R. & Urbini, G.  Modern
trend in plant design for the removal of nutritive substances
from sewage. In: Proc. of the 2nd Int. Symposium on
Advanced Wastewater Treatment, Krakow, Poland.
Ekama, G. A.  Theory, Design and Operation of Nutrient
Removal Activated Sludge Processes. Water Research
Commission (WRC), Pretoria, South Africa.
Ekama, G. A.  Biological nutrient removal (ch. 4.14) In:
Reference Module in Earth Systems and Environmental
Sciences – Treatise on Water Science. Elsevier, Amsterdam,
The Netherlands, pp. 409–526.
First received 23 November 2014; accepted in revised form 9 April 2015. Available online 21 April 2015
Water Science & Technology | 72.1 | 2015
Ekama, G. A. & Wentzel, M. C.  Denitrification kinetics in
biological N and P removal activated sludge systems treating
municipal wastewaters. Water Science and Technology 39 (6),
69–77.
Fajardo, C., Mosquera-Corral, A., Campos, J. L. & Méndez, R. 
Autotrophic denitrification with sulphide in a sequencing
batch reactor. Journal of Environmental Management
113, 552–556.
Henze, M.  Capabilities of nitrogen removal process from
wastewater. Water Science and Technology 23 (4–6),
669–679.
Luciano, A., Viotti, P., Mancini, G. & Torretta, V.  An
integrated wastewater treatment system using a BAS reactor
with biomass attached to tubular supports. Journal of
Environmental Management 113, 51–60.
Oh, J. & Silverstein, J. A.  Oxygen inhibition of activated
sludge denitrification. Water Research 33 (8), 1925–1937.
Plósz, B. G., Jobbágy, A. & Grady, C. P. Jr  Factors
influencing deterioration of denitrification by oxygen
entering an anoxic reactor through the surface. Water
Research 37 (4), 853–863.
Raboni, M., Torretta, V. & Urbini, G.  Influence of strong
diurnal variations in sewage quality on the performance of
biological denitrification in small community wastewater
treatment plants (WWTPs). Sustainability 5 (9), 3679–3689.
Raboni, M., Torretta, V., Viotti, P. & Urbini, G.  Calculating
specific denitrification rates in pre-denitrification by
assessing the influence of dissolved oxygen, sludge loading
and mixed-liquor recycle. Environmental Technology 35 (20),
2582–2588.
Sutton, P. M., Murphy, K. L. & Jank, B. E.  Nitrogen control: a
basis for design with activated sludge systems. In: Proc. of
IAWPR Conference, 18–20th August, Copenhagen, Denmark.
Tchobanoglous, G., Burton, F. L. & Stensel, H. D. 
Wastewater Engineering: Treatment and Reuse. 4th edn.
McGraw Hill, New York, USA.
Torretta, V., Ragazzi, M., Trulli, E., De Feo, G., Urbini, G., Raboni,
M. & Rada, E. C.  Assessment of biological kinetics in a
conventional municipal WWTP by means of the oxygen
uptake rate method. Sustainability 6 (4), 1833–1847.
US-EPA  Process Design Manual for Nitrogen Control. Water
Resource Center, Jefferson, MO, USA.
US-EPA  Nitrogen Control – Manual (EPA/625/R-93/010).
Office of Research and Development, Washington, DC, USA.
US-EPA  Nutrient Control Design Manual – State of
Technology Review Report (EPA/600/R-09/012). Watertown,
MA, USA.
US-EPA  Nutrient Control Design Manual (EPA/600/R-10/
100). Watertown, MA, USA.
Vismara, R.  Depuratione Biologica: Teoria e Processi
(Biological Treatment: Theory and Processes). 3rd edn.
Hoepli, Milano, Italy.