Aquacultural Engineering 34 (2006) 198–213

www.elsevier.com/locate/aqua-online

Comparative performance of fixed-film biological filters:

Application of reactor theory
Barnaby J. Watten *, Philip L. Sibrell
United States Geological Survey, Leetown Science Center, 11649 Leetown Road,
Kearneysville, WV 25430, USA

Received 20 December 2004; accepted 25 March 2005

Abstract

Nitrification is classified as a two-step consecutive reaction where R1 represents the rate of formation of the intermediate
product NO2-N and R2 represents the rate of formation of the final product NO3-N. The relative rates of R1 and R2 are
influenced by reactor type characterized hydraulically as plug-flow, plug-flow with dispersion and mixed-flow. We develop
substrate conversion models for fixed-film biofilters operating in the first-order kinetic regime based on application of
chemical reactor theory. Reactor type, inlet conditions and the biofilm kinetic constants Ki (h
1) are used to predict changes
in NH4-N, NO2-N, NO3-N and BOD5. The inhibiting effects of the latter on R1 and R2 were established based on the g
relation, e.g.:

N
Ki ½BOD5 
g¼ ¼ 1

Ki;max ½BOD5;max 

where BOD5,max is the concentration that causes nitrification to cease and N is a variable relating Ki to increasing BOD5.
Conversion models were incorporated in spreadsheet programs that provided steady-state concentrations of nitrogen and
BOD5 at several points in a recirculating aquaculture system operating with input values for fish feed rate, reactor volume,
microscreen performance, make-up and recirculating flow rates. When rate constants are standardized, spreadsheet use
demonstrates plug-flow reactors provide higher rates of R1 and R2 than mixed-flow reactors thereby reducing volume
requirements for target concentrations of NH4-N and NO2-N. The benefit provided by the plug-flow reactor varies with
hydraulic residence time t as well as the effective vessel dispersion number, D/mL. Both reactor types are capable of
providing net increases in NO2-N during treatment but the rate of decrease in the mixed-flow case falls well behind that
predicted for plug-flow operation. We show the potential for a positive net change in NO2-N increases with decreases in the
dimensionless ratios K2;ðR2 Þ =K1;ðR1 Þ and [NO2-N]/[NH4-N] and when the product K1;ðR1 Þ t provides low to moderate NH4-N
conversions. Maintaining high levels of the latter reduces the effective reactor utilization rate (%) defined here as (RNavg/
RNmax)100 where RNavg is the mean reactive nitrogen concentration ([NH4-N] + [NO2-N]) within the reactor, and RNmax
represents the feed concentration of the same. Low utilization rates provide a hedge against unexpected increases in
substrate loading and reduce water pumping requirements but force use of elevated reactor volumes. Further g effects on R1

* Corresponding author. Tel.: +1 304 724 4425; fax: +1 304 724 4428.
E-mail addresses: barnaby_watten@usgs.gov (B.J. Watten), philip_sibrell@usgs.gov (P.L. Sibrell).

0144-8609/$ – see front matter. Published by Elsevier B.V.

doi:10.1016/j.aquaeng.2005.03.006
 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213
and R2 can be reduced through use of a tanks-in-series versus a single mixed-flow reactor configuration and by improving the
solids removal efficiency of microscreen treatment.
Published by Elsevier B.V.
Keywords: Modeling; Nitrification; Aquaculture; Reactor theory; Performance
1. Introduction
Fixed-film biological filters are often used to
oxidize fish excretory products in recirculating
aquaculture systems (RAS) so as to reduce both
make-up water requirements and pollutant loads on
receiving streams (Liao and Mayo, 1974; Wheaton
et al., 1994; Shnel et al., 2002). Fixed-film treatment
also shows potential for improving discharge quality
in single pass fish culture applications (Losordo and
Westers, 1994; Bergheim and Brinker, 2003). In both
cases, biofilm development is encouraged within
reactors designed to provide the substrate, surface area
and hydrodynamic conditions necessary to maintain
activity at high biomass concentrations. Active
surfaces within the reactor consist of a matrix of
microorganisms and extra-cellular polymers that
allow for attachment to solid surfaces (Hagopian
and Riley, 1998). The composition, density and
thickness of the biofilm that develops, following an
initial conditioning period, is influenced by site
specific conditions including wastewater composition,
substrate flux, biofilm growth, decay and shear losses
(Rittmann and McCarty, 1980, 1981; Meunier and
Williamson, 1981). Biofilms in typical aquaculture
applications include autotrophic bacteria of the genus
Nitrosomonas and Nitrobacter responsible for the
conversion of NH4-N to NO3-N as well as hetero-
trophic bacteria capable of oxidizing organic matter,
e.g., the genus Acenitobacter (Bullock et al., 1993).
Heterotrophic bacteria exhibit a relatively high net cell
yield per unit mass substrate oxidized (Rittmann and
Snoeyink, 1984) which can, under high organic
loading conditions, lead to inhibition or displacement
of nitrifers from the biofilm matrix (Liao and Mayo,
1974; Ohashi et al., 1995; Satoh et al., 2000; Zhu and
Chen, 2001). Inhibition or displacement can result in
incomplete nitrification with an associated rise in
NH4-N and NO2-N above design limits established to
circumvent ammonia toxicity (Colt and Armstrong,
1981) and nitrite induced methemoglobinemia (Smith
199
and Williams, 1974; Manthe et al., 1984). Reactor
staging can be used to improve nitrification rates when
treating waters with high concentrations of organic
matter (Srna, 1975; Antonie, 1976; Miller and Libey,
1984; Watten et al., 1993). This configuration
promotes the development of biofilms adapted
specifically to wastewaters present within each
successive stage, e.g., heterotrophs in initial stages
and nitrifers in later stages. Additionally, staging can
be used to establish fluid residence time distributions
(RTD) that approximate plug-flow behavior known to
provide superior substrate conversion when reaction
rates are substrate limited—reaction orders > 0
(Levenspiel, 1972). Examples of fixed-film reactors
designed to operate in the plug-flow mode include
multi-stage RBC’s (Grady and Lim, 1980; Miller and
Libey, 1984), up-flow and down-flow submerged beds
(Haug and McCarty, 1972; Paller, 1992; Ridha and
Cruz, 2001), trickle filters (Rogers and Klemetson,
1985; Kruner and Rosenthal, 1983; Kamstra et al.,
1998) and fluidized beds (Summerfelt and Cleasby,
1993; Nam et al., 2000; Sandu et al., 2002). The latter
type is unique in that the solid phase (media with
attached biofilm) is mixed providing a homogenous
distribution of reactive surfaces and the potential for
providing effluent substrate concentrations that lie
below that required, on average, to maintain an active
biofilm (Rittmann, 1982). The solid phase used in the
alternative plug-flow filter types is fixed in position.
This requirement establishes a heterogeneous biofilm
along the longitudinal axis of the reactor (Haug and
McCarty, 1972; Nijhof, 1995). Mixed-flow reactors
are designed to operate with complete mixing in both
liquid and solid phases. Examples here include the
moving bed biofilter (Rusten et al., 1998; Zhu and
Chen, 1999), the rotating biodrum (Rogers and
Klemetson, 1985; Wheaton et al., 1994) and single
stage RBC’s. Mixed-flow reactors are preferred when
reaction rates are not substrate limited—reaction
order = 0 (Levenspiel, 1972). Biofilm reactor type
must be chosen carefully given its effect on
200 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213
performance as measured by required volume,
pumping requirements and changes in water chemistry
(Rittmann, 1982; Levenspiel, 1972; Weber, 1972). We
quantify these effects following development of
performance algorithms that allow standardization
of certain steady-state biofilm kinetic parameters.
Model use focuses on both organic effects and
incomplete nitrification given the sensitivity of fish
to NO2-N at relatively low concentrations.
2. Model development
2.1. Nitrification
Nitrification within the biofilm is classified as a
two-step consecutive reaction where R1 represents the
rate of formation of the intermediate product NO2-N
and R2 represents the rate of formation of the final
product NO3-N. The stoichiometry of the reaction has
been described as (Haug and McCarty, 1972):
R1 : 55NH4 þ þ 5CO2 þ 76O2
! C6 H7 O2 N þ 54NO2
þ 52H2 O þ 109Hþ
(1)
R2 : 400NO2
þ 5CO2 þNH4 þ þ 195O2 þ2H2 O
! C6 H7 O2 N þ 400NO3
þ Hþ (2)
where C6H7O2N represents the yield (mass) of bac-
terial cells. The reactions above are further classified
as being heterogeneous, i.e., fluid and solid phases are
present and necessary for the reactions to proceed. In
both cases, reaction kinetics follow the mixed-order
Monod relationship (Benefield and Randall, 1980):
dS km xS
¼ (3)
dt Ks þ S
where (dS/dt) is the overall substrate utilization rate,
mass volume
1 time
1, km the maximum rate of sub-
strate utilization per unit of attached biomass, time
1,
S the substrate concentration, mass volume
1, Ks the
saturation constant which has a value equal to the
substrate concentration when (dS/dt)/x = 0.5km, and x
is the active biomass concentration, mass volume
1.
Typical values for Ks;NH4 -N and Ks;NO2 -N are, respec-
tively, 0.5–2.0 mg/L (Benefield and Randall, 1980)
and 2.7 mg/L (Rittmann and Snoeyink, 1984;
T = 24 8C). When S is much greater than Ks, Equation
(3) reduces to an expression that is zero-order with
respect to S. When S is limiting and smaller than Ks, as
in typical salmonid aquaculture applications Equation
(3) reduces to an expression that is first-order with
respect to S. Here, the rate expressions for each
nitrogen component of the reaction can be expressed
as (Levenspiel, 1972; Srna and Baggaley, 1975;
Weatherley, 1984):
1 dMNH4 -N
rNH4 -N ¼ ¼ K1 ½NH4 -N (4)
V dt
1 dMNO2 -N
rNO2 -N ¼ ¼ K1 ½NH4 -N
K2 ½NO2 -N
V dt
(5)
1 dMNO3 -N
rNO3 -N ¼ ¼ K2 ½NO2 -N (6)
V dt
where M is the mass substrate reacted, V reactor
volume and K1 and K2 are the empirically derived
first-order reaction rate constants (time
1) that repre-
sent the product of a substrate utilization rate constant
and the active microbial mass per unit volume of the
reactor. Chemical reaction kinetics, as described in the
equations above, have been used with success to
follow the progress of nitrification within fixed-film
reactors (O’Shaughnessy et al., 1984; Srna and Bag-
galey, 1975; Weatherley, 1982, 1984; Weatherley
et al., 1993; Watten et al., 1993, 1998). Values of
K1 and the ratio K2/K1 will vary with reactor type and
application. Srna and Baggaley (1975) established rate
constants for a marine subgravel filter system operat-
ing with an inlet NH4-N of 0.1–0.7 mg/L. Tempera-
ture, pH and sulfide ion (0.03 mg/L) affected K1 and
K2 but small changes in salinity did not (5–8 ppt). K1
(h
1) rose with increasing pH whereas K2 (h
1) fell as
described in the relations: K1 = 0.230 (pH)
1.6
(7.4 < pH < 7.8); K2 =
0.053 (pH) +0.531
(6.6 < pH < 8.6). O’Shaughnessy et al. (1984) pre-
sent rate constants for a horizontal plug-flow reactor
25 m in length. K1 over the NH4-N range of 1.0–
1.5 mg/L averaged just 0.039 h
1 reflecting the low
specific surface area provided by the reactor design
(156 m2/m3). K2/K1 averaged 1.91. Weatherley (1984)
correlated K1 and K2 with pH based on Srna and
Baggaley (1975), temperature as described by an
Arrhenius type relationship and dissolved oxygen
(DO) following the observations of Knowles et al.
 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213 201

(1965). A rate modifying factor for the latter was used where t is expressed in hours. Nitrification is, for all
to describe a linear reduction in K1 with decreases in practical purposes, a unimolecular type reaction
DO within the threshold range 0.7–2.0 mg/L and for (Equation (1) and (2)). Therefore, the following nitro-
K2 within the threshold range 0.7–4.0 mg/L. Predicted gen balance holds:
Ki values agreed well with values established for a
subgravel filter during model verification runs. Here, ½NH4
Nin ¼ ½NH4 -Nout þ ½NO2 -Nout þ ½NO3 -Nout
K1 and K2 averaged 0.430 and 0.100 h
1, respectively, (9)
providing a K2/K1 ratio of 0.233. Ebeling (2000)
correlated K1 with hydraulic loading (HL) during an which leads to a solution for [NO3-N]out (Weber,
evaluation of a commercial bubble-wash bead filter 1972):
(T = 21–23 8C). K1 increased with HL from a low of

12.0 h
1 at HL = 13.2 L/min to a high of 35.5 h
1 at K1 e
K2 t K2 e
K1 t
½NO3 -Nout ¼ ½NH4 -Nin 1 þ

HL = 80.6 L/min. Brazil (2005) reported rate con- K2
K1 K2
K1
stants for a commercial scale RBC operating at a (10)
temperature of 28 8C. In this case, K1 averaged
2.4 h
1 with a K2/K1 average of 4.5. Watten et al. Equations (8)–(10) assume NO2-N and NO3-N are
(1993) reported a mean K1 and K2/K1 ratio of 13.9 h
1, not present in the influent, an unusual case in salmonid
and 1.6 for a fluidized bed of plastic beads (960 m2/ aquaculture operations. Correcting Equations (8)–(10)
m3) operating with an inlet NH4-N of 0.2–1.5 mg/L for inlet concentrations above zero results in:
(14.3 8C). Concurrently, K1 and the ratio K2/K1 varied
with location in a second reactor treatment of equal ½NO2 -Nout
volume that consisted of three fluidized bed compo- K1 ½NH4 -Nin
K1 t
K2 t
nents operated in series—K1 and the ratio K2/K1 were ¼ ðe
e Þ þ ½NO2 -Nin e
K2 t
K2
K1
15.8 h
1 and 2.8 for stage 1, 20.3 h
1 and 1.5 for stage (11)
2, and 12.8 h
1 and 3.1 for stage 3. Staging effects on
Ki values have also been reported for RBC applica-

tions (Brazil, 2005) reflecting changes in component K1 e
K2 t K2 e
K1 t
½NO3 -Nout ¼ ½NH4 -Nin 1 þ

feed composition and substrate loading. Clearly, Ki K2
K1 K2
K1
values vary widely and must therefore be established þ ½NO2 -Nin ð1
e
K2 t Þ þ ½NO3 -Nin
for a specific set of operating conditions either empiri- (12)
cally or through application of theoretical models like
those of Rittmann (1982) and Rittmann and McCarty The rate constant K1 is calculated based on
(1980, 1981). Equation (7) given NH4-N conversion, t and the
relation:
2.2. Plug-flow conversion

½NH4 -Nout
Integration of Equations (4) and (5) gives effluent K1 t ¼ ln (13)
½NH4 -Nin
NH4-N, NO2-N and NO3-N concentrations versus
reactor retention time t when the influent NH4-N The rate constant K2 is solved for by trial and error
concentration is known. In the plug-flow reactor case, given [NH4-N]in, [NO2-N]in, [NO2-N]out, t, Equation
we have (Levenspiel, 1972; Srna and Baggaley, 1975; (11), and a corresponding K1 from Equation (13).
Weatherley, 1982, 1984):
2.3. Mixed-flow conversion
½NH4 -Nout ¼ ½NH4 -Nin e
K1 t (7)
Integration of the rate expressions rNH3 -N (Equation
(4)) and rNO2 -N (Equation (5)) for the mixed-flow
K1 ½NH4 -Nin
K1 t reactor case gives alternate conversion models for each
½NO2 -Nout ¼ ðe
e
K2 t Þ (8)
K2
K1 of the nitrogen components of interest (Levenspiel,
202 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213

1972; Weber, 1972): the expression (Levenspiel, 1979):


2

1 s2 D D
mL=D
½NH4 -Nout ¼ ½NH4 -Nin (14) ¼2
2 1
e (20)
1 þ K1 t t2 mL mL

 where D/mL is dimensionless. D/mL varies from 1 for
K1 t
½NO2 -Nout ¼ ½NH4 -Nin ideal mixed-flow to 0 for ideal plug-flow. When non-
ð1 þ K1 tÞð1 þ K2 tÞ
ideal behavior occurs, the effects of D/mL on substrate
(15) conversion can be modeled as a variable number n of

 mixed-flow reactors operating in series where n, under
K1 K2 t 2 certain conditions, is correlated with D/mL following
½NO3 -Nout ¼ ½NH4 -Nin
ð1 þ K1 tÞð1 þ K2 tÞ the relation 1/n = 2 (D/mL) (Levenspiel, 1979). When
(16) first-order kinetics apply the model below gives
conversion of NH4-N as a function of n, t and K1
Correcting for the effects of inlet NO2-N and (Levenspiel, 1972):
NO3-N results in the expressions:


 1
K1 t ½NH4 -Nout ¼ ½NH4 -Nin (21)
½NO2 -Nout ¼ ½NH4 -Nin ð1 þ K1 ti Þn
ð1 þ K1 tÞð1 þ K2 tÞ

 Here, we assume individual reactor components of the
1
þ ½NO2 -Nin (17) series have a common volume and residence time, ti.
1 þ K2 t
When compared to an ideal plug-flow case, the rela-

 tive volume requirements of the tank-in-series reactor
K1 K2 t 2 reduces to the following at a fixed rate of conversion
½NO3 -Nout ¼ ½NH4 -Nin
ð1 þ K1 tÞð1 þ K2 tÞ (Levenspiel, 1972):


1
þ ½NO2 -Nin 1
þ ½NO3 -Nin Vseries K1 t
1 þ K2 t ¼1þ (22)
Vplug-flow 2n
(18)
where V is total volume and t is total retention time.
Here, K1 is calculated given changes in [NH4-N], t, Note that plug-flow behavior is approached at a
and the relation below derived from Equation (14): decreasing rate with increasing values of n.


1
K1 t ¼
1 (19) 2.5. Oxidation of organic matter
½NH4 -Nout =½NH4 -Nin

K2 is solved for by trial and error given [NH4-N]in,
[NO2-N]in, [NO2-N]out, t, Equation (17) and a corre-
sponding K1 from Equation (19).
2.4. Tanks-in-series conversion
The conversion models presented thus far assume
ideal plug-flow or mixed-flow reactor behavior. Ideal
reactor behavior is difficult to achieve, particularly in
the plug-flow case, given the sensitivity of RTD to
system geometry. The vessel dispersion coefficient, D/
mL, is used to identify reactor type given the measured
mean residence time, t, and variance, s2, of a
conservative tracer. These variables are related with
The rate constants K1 and K2 used to establish
reactor performance will vary with operating condi-
tions including the feed concentration of degradable
organic matter that stimulates heterotrophic bacterial
growth (Bovendeur et al., 1990; Ohashi et al., 1995;
Zhu and Chen, 2001). The stoichiometry of the
oxidation reaction associated with organic matter has
been expressed as (Benefield and Randall, 1980):
C5 H12 þ 8O2 ! 5CO2 þ 6H2 O (23)
Although reaction kinetics adhere to the mixed-
order Monod relationship (Equation (3)), S in
salmonid RAS applications (e.g., BOD5) is typically
much smaller than Ks resulting in kinetics that are
 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213 203

again first-order with respect to S, i.e.

1 dMBOD5
rBOD5 ¼ ¼ K3 ½BOD5  (24)
V dt
where K3 is the BOD5 reaction rate constant (time
1)
or treatability factor that represents the product of a
substrate utilization rate constant and the active micro-
bial mass per unit volume of the reactor (Benefield and
Randall, 1980). K3 is related to operating conditions
that include for example, in the trickle filter case,
temperature, packing type, packing depth, surface
hydraulic loading and waste source (Eckenfelder,
1961). Additional variables of importance are dis-
solved oxygen and pH (Benefield and Randall,
1980). The first-order conversion models developed
Fig. 1. Effect of the BOD poisoning coefficient N on the response of
for plug-flow (Equation (7)), mixed-flow (Equation the ratio K1/K1,max to increases in the ratio BOD5/BODmax.
(14)) and tanks-in-series (Equation (21)) reactors
apply here when [NH4-N] and K1 are replaced with
[BOD5] and K3. For example, in the mixed-flow case case, however, a two-step calculation procedure can be
we have: used—[BOD5]out is established based on Equation

 (25), then is used in a second step to adjust both K1 and
1 K2 for inhibition (Equation (26)). Adjusted Ki values
½BOD5 out ¼ ½BOD5 in (25)
1 þ K3 t are then used to predict effluent concentrations of
NH4-N, NO2-N and NO3-N as described earlier
We correct K1 and K2 for the inhibiting effects of
(Equations (14)–(18)). The mixed-flow calculation
BOD5 using a model form (Levenspiel, 1979)
procedure just described, when combined with the
developed to describe product poisoning in microbial
tanks-in-series approach, allows conversion rates to be
fermentation systems. In our case, K1 or K2 established
identified for the plug-flow (with dispersion) reactor
without inhibition (Ki,max) is adjusted given the BOD5
operating with g < 1. Examples include the trickle
concentration that causes nitrification to cease
filter and submerged bed filter. Here, concentrations of
(BOD5,max), the steady-state concentration of BOD5
BOD5, NH4-N, NO2-N and NO3-N calculated in
within the reactor, and a value for N, a coefficient that
effluent of the first reactor stage of a selected number
characterizes the response of Ki values to increasing
of stages (n) are used as the influent condition for the
BOD5 concentrations, i.e.:
second reactor stage in the series. This stage-by-stage

N calculation procedure is repeated until the effluent
Ki ½BOD5 
g¼ ¼ 1
(26) condition has been identified for the last tank in the
Ki;max ½BOD5;max 
series. This procedure also allows Ki,max to be
If g < 1, the apparent values of Ki established based adjusted, when necessary, for changes in biofilm
on Equation (19) represent the product gKi,max. Fig. 1 characteristics that may occur independently of the
gives a plot of K1/K1,max versus BOD5/BOD5,max for effects of degradable organics (g).
selected values of N. Note that initial reductions in K1
with increasing BOD5 are accelerated as the coeffi- 2.6. Spreadsheet program simulating RAS water
cient N increases. When BOD5 oxidation occurs and chemistry
g < 1, application of Equation (26) precludes use of
the steady-state conversion models developed for the We developed an Excel spreadsheet to carry out the
plug-flow and tanks-in-series reactors given that calculations required to establish steady-state con-
BOD5, and hence Ki/Ki,max, will vary along the centrations of NH4-N, NO2-N and NO3-N within a
longitudinal axis of the reactors. In the mixed-flow salmonid RAS versus selected operating conditions.
204 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213
Fig. 2. Spreadsheet model components used to simulate the effect of reactor type on water quality dynamics in an RAS. The letters (a–j)
represent model prediction points for the variables BOD5, NH4-N, NO2-N and NO3-N.
Model inputs included reactor type (plug-flow or
mixed-flow), K1, the ratio K2/K1, system make-up flow
rate (Qmf, m3/h), recirculating flow rate (Qrf, m3/h),
make-up flow composition ([NH4-N]mf, [NO2-N]mf,
[NO3-N]mf, mg/L), reactor volume (V, m3) and fish
feed rate (kg/d). The latter was used to calculate NH4-
N loading based on the ratio 29 g/kg feed which
represents a typical value for salmonids (Speece,
1973; Liao and Mayo, 1974). Net changes in nitrogen
components were established using the reactor
specific conversion models described earlier (either
Equations (7)–(12) or (14)–(18)) in combination with
a material balance. For example, in the case of NH4-N
the following balance holds:
Qmf ½NH4 -Nmf þ MNH4
N ¼ Qof ½NH4 -Nof
þ ðQmf þ Qrf Þð½NH4 -Nin
½NH4 -Nout Þ
where in this case NH4-N concentrations are
expressed as g/m3, Qof is a system overflow rate
(Qof = Qmf, m3/h) and MNH4 -N is the mass of NH4-
N produced by the fish (g/h). The spreadsheet was
designed to allow system operation without recircula-
tion, if desired, so as to simulate single pass or post
treatment applications of the biofilm reactors, i.e.,
Qrf = 0. When recirculation is employed (Qrf > 0),
an iteration procedure and check for convergence
was required given the rise in reactant concentrations
that occur with time when reactant conversion per pass
is less than 100%. Organic inhibition was not
addressed here but was in a similar spreadsheet
designed to simulate the effects of a microscreen
BOD5 removal step and reactor staging. Major com-
ponents of the system modeled are given in Fig. 2. Net
BOD5 input was taken as the product of fish feed
rate (kg/d), 1
E, and the ratio g biosolids/kg feed,
where E represents the biosolids removal efficiency
of the microscreen as a decimal fraction. We used a
net biosolids (BOD5) production rate of 400 g/kg
feed based on the observations of Speece (1973)
and Liao and Mayo (1974). BOD5 was tracked through
each RAS component with conversion accounted for
in the primary mixed-flow reactor and tanks-in-series
reactors (Equation (25)). Reactor specific BOD5
concentrations were used to correct K1 and K2 for
inhibition effects given a selected value for N
(27) (Equation (26)). Additional program inputs included
reactor volumes and reactor configuration—either a
single mixed-flow reactor, a tanks-in-series reactor
(six stages) or a combination of the both types as
illustrated in Fig. 2. In each of the three spreadsheet
programs we assume dissolved oxygen, CO2, tem-
perature, pH and alkalinity are maintained within
desirable ranges by RAS system components not
included in our modeling exercise and that the unit
processes/operations used here do not influence the
reactants of interest. Further, and following the RAS
models of Weatherley (1982, 1984) and Weatherley
et al. (1993), we do not address conversion of BOD5,
NH4-N or NO2-N that may occur outside of the
 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213 205

reactors in the bulk recirculating flow or on piping or

tank surfaces.

3. Model application

We used the spreadsheets to establish reactor

sensitivity to certain operating conditions and to
compare reactor performance. Fig. 3 demonstrates a
typical response of reactor output to changes in
hydraulic residence time t. Here, the NH4-N present in
a plug-flow reactors effluent drops rapidly following a
first-order conversion (R1) to the intermediate product
NO2-N (Equations (1) and (7)). Concurrently, NO2-N
rises then falls as its own rate of conversion to NO3-N
begins to exceed R1. This response is similar to that
observed in experimental biofilm reactors (Ebeling,
2000; O’Shaughnessy et al., 1984; Srna and Baggaley,
1975; Zhu and Chen, 1999) and demonstrates a
declining rate of return on investments in reactor
volume or t, particularly with regards to R1. This latter
point is illustrated in Fig. 4. Here, NH4-N conversion
is plotted versus reactor retention time for both plug-
flow and mixed-flow reactors operating with a
common K1 of 3 or 14.5 h
1. The K1 values used
are similar to those reported for a commercial scale
RBC (Brazil, 2005) and a pilot scale fluidized bed
(Watten et al., 1993). Note that in accordance with
theory, the plug-flow reactor provides higher levels of
conversion given that S (NH4-N) in the mixed-flow
Fig. 4. Effect of reactor type on the conversion of [NH4-N] vs.
hydraulic retention time when K1 is fixed at 3 or 14.5 h
1.
case is uniformly low throughout the reactor and equal
to the reactor’s effluent concentration. The disparity in
conversion rates will decrease as reactor geometry is
altered so as to lower the working vessel dispersion
number (D/mL) from the mixed-flow case (1) towards
the plug-flow case (0), e.g., by adopting the tanks-in-
series configuration. This effect is illustrated in Fig. 5
where the percent reduction in NH4-N conversion is
plotted versus K1t and D/mL. The reduction in
conversion in this example drops from a high of
about 23% with D/mL = 1 to a low of about 3% with D/
mL = 0.07. The latter corresponds to an RTD provided
by seven mixed-flow reactors operated in series. Note

Fig. 3. Effluent composition of a plug-flow biofilm reactor vs.

hydraulic retention time when inlet [NH4-N] = 1 mg/L. Line A Fig. 5. Effect of the product K1t on the relative rate of NH4-N
represents reactive nitrogen (RN) concentration, i.e., NH4- conversion (%) at selected levels of hydraulic dispersion (D/mL)
N + NO2-N. Line B represents RNmax. within the reactor.
206 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213
Fig. 6. Effect of reactor type and volume on steady-state concen-
trations of NH4-N and NO2-N in the rearing unit component of an
RAS. Arrows indicate volume requirements for a selected [NH4-N]
of 1 mg/L and a selected [NO2-N] of 0.2 mg/L.
also that the benefit provided by a plug-flow reactor
rises and then falls as K1t increases. This response is
explained by the converging levels of ammonia removal
shown previously in Fig. 4. Fig. 6 gives the net effect of
reactor type on volume requirements for a selected set
of RAS operating conditions including feed rate (60 kg/
d), total recirculating flow (99 m3/h) and make-up flow
(1 m3/h). NH4-N within the rearing unit declines at a
decreasing rate with increasing reactor volume in both
plug-flow and mixed-flow modes but NH4-N is always
lower in the plug-flow case. The effect of this relation on
reactor volume requirements will vary, as shown, with
the target NH4-N concentration, e.g., selecting a
target of 1 mg/L results in volume requirements of 9
and 18.5 m3 for the plug-flow and mixed-flow reactors,
respectively.
Rational design procedures for reactor sizing must
also include an analysis of NO2-N given linked effects
on fish stress and mortality (Colt and Armstrong,
1981; Manthe et al., 1984; Nijhof and Klapwijk,
1995). Liao and Mayo (1974) observed a rapid rise in
mortality of test salmonids when NO2-N levels rose
above 0.15 mg/L as a result of incomplete nitrification
in both trickle and submerged bed type biofilters.
Smith and Williams (1974) report exposure of
salmonids to 0.15 mg/L NO2-N for just 48 h results
in a 75% conversion of heme iron in hemoglobin from
the ferrous to the ferric state—methemoglobin. This
condition greatly reduces the ability of the blood to
transport oxygen and should therefore be avoided. In
those RAS applications that maintain very low make-
up-flow rates it may be feasible to negate the effect of
the NO2-N on hemoglobin by maintaining high Cl

concentrations in solution as practiced in pond
production of channel catfish (Tucker, 1985). Here,
the objective is to maintain a molar ratio of NO2/Cl of
0.25. This condition provides an acceptable compro-
mise between the cost/logistics of applying NaCl or
CaCl2 and the degree of conversion of hemoglobin to
methemoglobin. Fig. 6 also gives the rearing unit
NO2-N concentrations that correspond with the
changes in reactor volume. As with NH4-N, reactor
volume requirements for the target NO2-N level
(0.2 mg/L) are substantially lower for the plug-flow
versus the mixed-flow reactor application; 17 m3
versus 26 m3. Fig. 7 shows that NO2-N rises then falls
with increases in t in both reactor types but the rate of
decrease in the mixed-flow case falls well behind that
predicted for plug-flow operation. It is interesting to
note that K1 and the ratio K2/K1 influence the time at
which NO2-N is maximized (tmax), i.e., through
differentiation of the [NO2-N]out functions (Equations
(8) and (15)) and setting d[NO2-N]out/dt = 0, we
obtain for plug-flow (Levenspiel, 1972):
lnðK2 =K1 Þ
tmax;NO2 -N ¼ (28)
K2
K1
and for mixed-flow
1
tmax;NO2 -N ¼ pffiffiffiffiffiffiffiffiffiffiffi (29)
K2 K1
The [NO2-N]out corresponding to tmax;NO2 -N can be
derived by combining Equations (8) and (28) in the
plug-flow case:

K2 =ðK2
K1 Þ
K1
½NO2 -Nmax ¼ ½NH4 -Nin (30)
K2
 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213
Fig. 7. Effect of the ratio K1/K2 and hydraulic retention time on a
plug-flow and mixed-flow reactor’s effluent [NO2-N] when K1 is
fixed at 14.5 h
1and influent [NO2-N] is 0.2 mg/L.
and by combining Equations (15) and (29) in the
mixed-flow case:
1 N]in/[NH4-N]in. This relationship is in agreement
½NO2 -Nout;max ¼ ½NH4 -Nin pffiffiffiffiffiffiffiffiffiffiffiffiffi (31)
½ K2 =K1 þ 12
The equations above were developed with the
caveat [NO2-N]in = 0. Inspection reveals (1) as the
ratio K2/K1 tends toward 0, both tmax and [NO2-
N]out,max increase as illustrated in Fig. 7, (2) tmax
plug-flow is less than tmax mixed-flow and (3) plug-
flow [NO2-N]out,max is greater than mixed-flow
[NO2-N]out,max. Fig. 8 shows that when [NO2-N]in
is greater than 0, [NO2-N]out,max increases while tmax
is shifted towards lower values. Fig. 8 also shows the
ratio [NO2-N]out,max/[NO2-N]in declines as [NO2-
N]in rises in response, for example, to a reactor
operating with low conversion rates. Here, net
increases in NO2-N across the reactor are masked
by high ambient NO2-N levels and low NH4-N
conversions. Fig. 9 illustrates this effect, with K1/
207
Fig. 8. Effect of influent [NO2-N] on the peak [NO2-N] in effluent of
a plug-flow and mixed-flow reactor.
K2 = 0.75, as we show the ratio [NO2-N]out/[NO2-
N]in is dependent on the dimensionless ratio [NO2-
with the observations of Srna and Baggaley (1975)
and Nijhof and Klapwijk (1995). Fig. 10 shows that
[NO2-N]out drops below an example inlet NO2-N of
0.2 mg/L only when reactor volume provides for
high levels of NH4-N conversion (R1), i.e., with
operating K2/K1 ratios of 0.50, 0.75, 0.99 and 1.50
the corresponding NH4-N conversion requirements
are 97, 94, 88 and 77%, respectively. Surprisingly,
this same figure demonstrates that when K2/K1 is
fixed, mixed-flow reactors have a lower NH4-N
conversion requirement for the target [NO2-N]out
than plug-flow reactors—73% versus 88%. However,
following Equations (7) and (14), the reactor
retention time t required to achieve the target
conversion level (R1) with the mixed-flow reactor
is still greater then that associated with plug-flow
operation, in this case 30.3 min versus 8.8 min. In
208 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213
Fig. 9. Effect of K1t and the ratio [NO2-N]in/[NH4-N]in on the ratio [NO2-N]out/[NO2-N]in when operating a plug-flow reactor.
both reactor types, the high NH4-N conversion rates
required to establish the condition [NO2-N]out/[NO2-
N]in  1.0 lead to a substantial degree of under-
utilization of the biofilm substrate conversion
Fig. 10. Effect of the ratio K1/K2 and ammonia conversion (%) on
effluent [NO2-N] when operating a plug-flow reactor with an
influent [NO2-N] of 0.2 mg/L.
potential (g N/(m3 d)). For example, line A, Fig. 3,
gives the sum of NH4-N and NO2-N versus location
along the longitudinal axis of a plug-flow reactor
constructed with a uniform cross-sectional area. We
define the sum of these two nitrogen components as
reactive nitrogen (RN) given their combined or
individual abilities to sustain nitrification reactions.
Here, the microbes within the reactor work at a rate
that is fixed, in the first-order regime, by a mean RN
concentration that lies well below inlet levels
(RNmax). The effective RN concentration (RNavg)
can be calculated by applying the mean value
theorem of integral calculus given functions describ-
ing either the individual or combined (RN) nitrogen
component curves. Considering the combined func-
tion, we know from our material balance (Equation
(3)) that at any time t:
RN ¼ RNmax
½NO3 -Nout (32)
Then Equation (8), modified so as to reflect the
inverted mirror image relationship between RN and
NO3-N (Fig. 3), gives RN directly as a function of
[NH4-N]in, t, K1 and K2:


K1 e
K2 t K2 e
K1 t
½RNout ¼ ½NH4 -Nin 1
1þ

K2
K1 K2
K1
(33)
Applying the mean value theorem of integral
calculus to Equation (33) gives RNavg as a function of
 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213 209

the same variables:

RNavg ¼ ½NH4 -Nin


K1
K2 ðK2
K1 Þ ðe

K2 t

1Þ
K1 ðKK22
K1 Þ ðe
K1 t
1Þ

t
(34)

A somewhat different approach was used to

establish RNavg for the mixed-flow reactor. Here,
we start with [NO3-N]out based on Equations (16).
Recognizing that in this type of reactor RNout = RNavg
and considering Equation (32), we obtain RNavg as a
function of the variables of interest without applica-
tion of the mean value theorem:


K1 K2 t2
RNavg ¼ ½NH4 -Nin 1
(35)
ð1 þ K1 tÞð1 þ K2 tÞ
The RNavg calculated for a selected reactor type or
set of operating conditions is compared to RNmax to
establish the effective reactor utilization rate:


RNavg
Reactor utilization rate  ð%Þ ¼ 100
RNmax
(36)
Fig. 11. Effect of reactor type, K2/K1 and K1t on reactor utilization
The utilization rate then represents the degree to rate (%) = (RNavg/RNmax)  100.
which the biofilm present within the reactor is
saturated with RN relative to that present in the
reactor’s feed. Fig. 11 shows utilization rates drop Our modeling also suggests that high reactor utilization
with increases in K1t (or NH4-N conversion, rates are indicative of net increases in [NO2-N] across
Equation (7)) and rise as the ratio K2/K1 tends the reactor (Figs. 7–10) that may cause [NO2-N] to rise
toward 0. For example, with a K2/K1 = 0.75, K1t above species-specific standards. Therefore the design
values providing NH4-N conversions of 37 and 95% approach used to size the reactor should consider the
(plug-flow) result in reactor utilization rates of 97.8 interrelationships between conversion rate, pumping
and 60.8%, respectively. Reactor utilization rates are requirements, reactor capital, water chemistry and fish
somewhat lower in the mixed-flow versus the plug-flow health.
reactor when compared at a common K1t value Simulation runs discussed thus far have not
(Fig. 11). With K1t = 2.33 and K2/K1 = 0.75, for addressed the effects of organics (BOD5) on reactor
example, reactor utilization rates for the plug-flow performance. Fig. 12 gives the NH4-N and NO2-N
and mixed-flow reactors are 72.7 and 55.5%, respec- concentrations predicted for the rearing unit compo-
tively. Selecting a K1t with either reactor type that nent of the RAS defined in Fig. 2 when operating with
provides a high reactor utilization rate can minimize either of two reactor configurations—a single mixed-
reactor volumes needed to handle average daily loads flow reactor (3 m3 volume) or a tanks-in-series reactor
but (1) it will not provide the reserve capacity available (3 m3 volume). The latter includes a primary mixed-
with low utilization rates that are useful in hedging flow reactor (0.75 m3 volume) followed by six
against diel or unexpected increases in substrate loading secondary mixed-flow reactors (2.25 m3 total volume)
and (2) it will increase water pumping requirements. simulating plug-flow operation with a homogeneous
210 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213
Fig. 12. Effect of feed rate and reactor type on [NO2-N] and [NH4-
N] in the fish tank component of an RAS.
solid phase, e.g., a fluidized bed. The K3 value required
for BOD5 conversion estimates (6.3 h
1; Equation
(25)) was taken from Watten et al. (1993). The product-
poisoning coefficient N (Equation (26)) was fixed at a
value of 5 to explore a case of high initial sensitivity to
BOD5 (Fig. 1). BODmax (Equation (26)) was set at
20 mg/L, for example purposes only, based on the
observations of Antonie (1976) and Weng and Molof
(1974). The make-up flow requirement of the RAS in all
cases (0.1 m3/h) represented 1% of the total flow
directed through the rearing unit (10 m3/h). The
removal of suspended solids across the microscreen
was initially held at 60% (E = 0.6). This lies within the
range of E = 0.21
0.68 reported for RAS applications
by Bergheim and Brinker (2003) and Twarowska et al.
(1997). Fish feed rates were varied from 2 to 12 kg/d
which correspond to a change in dissolved oxygen
across the rearing unit of between 2.1 and 12.5 mg/L
assuming an oxygen demand ratio of 0.25 kg/kg feed
(Losordo and Westers, 1994). Model predicted con-
centrations of BOD5 in the microscreen effluent
(reactor feed) ranged from 2.0 to 12.2 mg/L in the
mixed-flow case and from 1.6 to 9.9 mg/L in the tanks-
in-series case. The latter provided lower NH4-N and
NO2-N concentrations than the mixed-flow reactor
(Fig. 12) in agreement with our earlier simulation
results established without BOD5 inhibition (Figs. 4–6).
The inhibition parameter g (Equation (26)), in the
mixed-flow case, dropped from 0.92 to 0.31 as feed rate
increased. In the tanks-in-series application, g varied
with position along the length of the reactor axis
reflecting stage-by-stage reductions in BOD5. Table 1
Fig. 13. Effect of microscreen efficiency and reactor type on [NO2-
N] and [NH4-N] in the fish tank component of an RAS.
gives g and BOD5 versus position as well as model
predicted NH4-N, NO2-N and NO3-N when operating
at feed rates of 2 and 12 kg/d. Note g increases, at
12 kg/d, from a low of 0.13 in the primary reactor to a
high of 0.61 in the last stage of the series reactor. An
initial rise then fall in NO2-N can be seen as water
moves from stage to stage following the response
predicted in Fig. 7. Further runs (Table 1) show
steady-state concentrations of NO2-N and NH4-N
drop significantly when the poisoning coefficient N is
reduced from 5 to 1, the latter establishing a linear
(negative) effect of BOD5 on g (Fig. 1), e.g., in the fish
tank effluent NO2-N and NH4-N drop from 0.55
to 0.14 and 1.99 to 1.55 mg/L, respectively. The g
factor here rises to a range of 0.61–0.91. Reducing
the microscreen efficiency to 30% (E = 0.3) acts to
lower g and elevate BOD5, NO2-N and NH4-N
(Table 1). Fig. 13 shows this effect on fish tank
effluent composition over the solids removal effi-
ciency range of 10–80%. In this case the N used to
define g was fixed at 3. Note that NH4-N and NO2-N
drop at a decreasing rate as solids removal efficiency
approaches 80%. The E-value required for a design
level of NH4-N and NO2-N are lower in the tanks-in-
series versus mixed-flow reactor configuration.
4. Summary
We develop substrate conversion models for fixed-
film biological filters operating in the first-order
kinetic regime based on application of chemical
Table 1
Model predicted water chemistry variables vs. sample point in an RAS operating under four test case conditions
Test case Variable RAS component

B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213

Fish tank Fish tank Micro-screen Primary reactor Tanks-in-series reactor
(in) (out) (out) (out) Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6
(out) (out) (out) (out) (out) (out)
2 kg/d; N = 5; E = 0.6 BOD5 0.3 3.6 1.6 1.1 0.9 0.7 0.6 0.5 0.4 0.3
g 0.75 0.79 0.83 0.86 0.89 0.91 0.92
NH4-N 0.02 0.26 0.26 0.14 0.10 0.07 0.05 0.03 0.02 0.02
NO2-N 0.02 0.02 0.02 0.06 0.06 0.06 0.05 0.04 0.03 0.02
NO3-N 23.87 23.87 23.87 23.94 23.98 24.02 24.05 24.07 24.09 24.11
12 kg/d; N = 5; E = 0.6 BOD5 1.9 21.9 9.9 6.7 5.4 4.4 3.5 2.9 2.3 1.9
g 0.13 0.21 0.29 0.38 0.46 0.54 0.61
NH4-N 0.54 1.99 1.99 1.75 1.58 1.37 1.14 0.92 0.72 0.54
NO2-N 0.55 0.55 0.55 0.65 0.71 0.75 0.75 0.71 0.64 0.55
NO3-N 141.65 141.65 141.65 141.79 141.90 142.07 142.29 142.55 142.83 143.10
12 kg/d; N = 1; E = 0.6 BOD5 1.9 21.9 9.9 6.7 5.4 4.4 3.5 2.9 2.3 1.9
g 0.67 0.73 0.78 0.82 0.86 0.88 0.91
NH4-N 0.10 1.55 1.55 0.92 0.66 0.47 0.33 0.23 0.15 0.11
NO2-N 0.14 0.14 0.14 0.38 0.40 0.37 0.31 0.25 0.15 0.11
NO3-N 143.27 143.27 143.27 143.67 143.90 144.13 144.33 144.50 144.63 144.73
12 kg/d; N = 1; E = 0.3 BOD5 3.3 23.2 17.3 11.7 9.5 7.7 6.2 5.0 4.1 3.3
g 0.41 0.53 0.62 0.69 0.75 0.80 0.84
NH4-N 0.17 1.62 1.62 1.13 0.89 0.67 0.49 0.35 0.25 0.17
NO2-N 0.21 0.21 0.21 0.43 0.47 0.46 0.42 0.35 0.28 0.21
NO3-N 143.14 143.14 143.14 143.41 143.61 143.83 144.06 144.27 144.44 144.58

211
212 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213
reactor theory (Equations (7), (11), (12), (14), (17),
(18), (25))—nitrification is treated as a two-step
consecutive reaction where R1 represents the rate of
formation of the intermediate product NO2-N and R2
represents the rate of formation of the final product
NO3-N. Spreadsheet programs incorporating the
models developed give concentrations of NH4-N,
NO2-N, NO3-N and BOD5 at several points within an
RAS for selected sets of operating conditions that
include the empirical rate constants Ki, reactor type
(plug-flow, mixed-flow, tanks-in-series), reactor
volume, make-up and recirculating flows, fish feed
rate and the efficiency of solids capture via micro-
screening. The inhibiting effects of BOD5 on
nitrification were established using the g relation
(Equation (26)) given an empirical value for BOD5,max
that causes nitrification to cease and a value for N, a
product poisoning coefficient linking Ki’s response to
increasing BOD5. Spreadsheet use demonstrates
reactor type must be chosen carefully given its effect
on performance as measured by required volume and
changes in water chemistry. Specifically, our models
show (1) NH4-N conversion (R1) rises at a decreasing
rate with increases in either K1 or t, (2) plug-flow
reactors provide higher rates of R1 than alternate
reactor types when K1t is fixed. The benefit rises with
K1t up to a maximum of 23% but decreases as the
effective vessel dispersion number D/mL tends toward
zero, (3) relatively high rates of R1 provided by plug-
flow reactors can serve to decrease required reactor
volumes for an RAS depending on target NH4-N
concentrations, (4) increases in K1t can cause [NO2-
N]out to rise above influent levels then fall as R2 begins
to exceed R1. The extent of the rise is related to the
relative rates of R1 and R2, reactor type (Equations (30)
and (31)) and influent reactive nitrogen concentrations
(Figs. 7–9), (5) high levels of NH4-N conversion can
be required to establish net decreases in NO2-N across
the reactor. Maintaining a high conversion rate leads to
a substantial degree of underutilization of the biofilm
substrate conversion potential (Equations (32)–(36))
but provides a hedge against unexpected increases in
substrate loading and (6) BOD5 inhibition effects on R1
and R2 are reduced as the poisoning coefficient N
(Equation (26)) decreases, when a tanks-in-series
reactor is used versus a single mixed-flow reactor, and
when the solids capture efficiency of the microscreen
component increases (Figs. 12 and 13).
References
Antonie, R.L., 1976. Fixed Biological Surfaces—Wastewater Treat-
ment. CRC Press Inc., Cleveland, Ohio.
Benefield, L.D., Randall, C.W., 1980. Biological Process Design for
Wastewater Treatment. Prentice Hall Inc., Englewood Cliffs,
New Jersey.
Bergheim, A., Brinker, A., 2003. Effluent treatment for flow through
systems and European environmental regulations. Aquacult.
Eng. 27, 61–77.
Bovendeur, J., Zwaga, A.B., Lobee, B.G.J., Blom, J.H., 1990. Fixed-
biofilm reactors in aquacultural water recycle systems: effects of
organic matter elimination on nitrification kinetics. Water Res.
24, 207–213.
Brazil, B.L., 2005. Performance and operation of a rotating biolo-
gical contactor system in a recirculating aquaculture system.
Aquacult. Eng., in press.
Bullock, G., Hankins, J., Heinen, J., Starliper, C., Teska, J., 1993.
Qualitative and quantitative bacteriological studies on a flui-
dized sand biofilter used in a semiclosed trout culture system.
Biological Report 17, U.S. Fish and Wildlife Service, Washing-
ton, DC.
Colt, J.E., Armstrong, D.A., 1981. Nitrogen toxicity to crustaceans,
fish and mollusks. In: Bio-Engineering Symposium for Fish
Culture, FCS Publication 1, American Fisheries Society,
Bethesda, Maryland, pp. 34–47.
Ebeling, J. M., 2000. Kinetic reaction rate analysis of nitrifying bead
filters in aquaculture. Ph.D. Dissertation, Department of Biolo-
gical Resources Engineering, University of Maryland.
Eckenfelder Jr., W.W., 1961. Trickling filter design and perfor-
mance. J. Sanitary Eng. Division, AACE SA6, 87–102.
Grady, L.C.P., Lim, H., 1980. Biological Wastewater Treatment:
Theory and Application. Marcel Dekker Inc., New York.
Hagopian, D.S., Riley, J.G., 1998. A closer look at the bacteriology
of nitrification. Aquacult. Eng. 18, 223–244.
Haug, R.T., McCarty, P.L., 1972. Nitrification with submerged
filters. J. Water Poll. Control Fed. 44, 2086–2102.
Kamstra, A., Van der Heul, J.W., Nijhof, M., 1998. Performance and
optimization of trickling filters on eel farms. Aquacult. Eng. 17,
175–192.
Knowles, G., Downing, A.L., Barrett, M.J., 1965. Determination
of kinetic constants for nitrifying bacteria in mixed culture,
with the aid of an electronic computer. J. Gen. Microbiol. 38,
263–278.
Kruner, G., Rosenthal, H., 1983. Efficiency of nitrification in trickling
filters using different substrates. Aquacult. Eng. 2, 49–67.
Levenspiel, O., 1972. Chemical Reactor Engineering. John Wiley
and Sons Inc., New York.
Levenspiel, O., 1979. The Chemical Reactor Omnibook. OSU
Bookstores Inc., Corvallis, OR.
Liao, P.B., Mayo, R.D., 1974. Intensified fish culture combining
water reconditioning with pollution abatement. Aquaculture 3,
61–85.
Losordo, T.M., Westers, H., 1994. System carrying capacity and
flow estimation. In: Timmons, M.B., Losordo, T.M. (Eds.),
Aquaculture Water Reuse Systems; Engineering Design and
Management. Developments in Aquaculture and Fisheries
 B.J. Watten, P.L. Sibrell / Aquacultural Engineering 34 (2006) 198–213
Science, vol. 27. Elsevier Science, Amsterdam, The Nether-
lands, pp. 9–60.
Manthe, D.P., Malone, R.F., Kumar, S., 1984. Limiting factors
associated with nitrification in closed blue crab shedding sys-
tems. Aquacult. Eng. 3, 119–140.
Meunier, A.D., Williamson, K.J., 1981. Packed bed biofilm reactors:
simplified model. J. Environ. Eng. Division, ASCE EE2, 307–
315.
Miller, G.E., Libey, G.S., 1984. Evaluation of a trickling biofilter in a
recirculating aquaculture system containing channel catfish.
Aquacult. Eng. 3, 39–57.
Nam, T.K., Timmons, M.B., Montemagno, C.D., Tsukuda, S.M.,
2000. Biofilm characteristics as affected by sand-size and loca-
tion in fluidized bed vessels. Aquacult. Eng. 22, 213–224.
Nijhof, M., 1995. Bacterial stratification and hydraulic loading
effects in a plug-flow model for nitrifying trickling filters applied
in recirculating fish culture systems. Aquaculture 134, 49–64.
Nijhof, M., Klapwijk, A., 1995. Diffusional transport mechanisms
and biofilm nitrification characteristics influencing nitrite levels
in nitrifying trickling filter effluents. Water Res. 29, 2287–2292.
O’Shaughnessy, J.C., Blanc, F.C., Barker, T.G., 1984. Fixed film
nitrification kinetics. In: Second International Conference on
Fixed Film Processes, Pittsburgh, Pennsylvania, pp. 1389–1403.
Ohashi, A., Viraj de Silva, D.G., Mobarry, B., Manem, J.A., Stahl,
D.A., Rittmann, B.E., 1995. Influence of substrate C/N ratio on
the structure of multi-species biofilms consisting of nitrifers and
heterotrophs. Water Sci. Tech. 32, 75–84.
Paller, M.H., 1992. An analytical model for predicting the carrying
capacity of submerged biofilters used in aquaculture. J. Appl.
Aquacult. 1, 1–25.
Ridha, M.T., Cruz, E.M., 2001. Effect of biofilter media on water
quality and biological performance of the Nile tilapia Oreo-
chromis niloticus L. reared in a simple recirculating system.
Aquacult. Eng. 24, 157–166.
Rittmann, B.E., 1982. Comparative performance of biofilm reactor
types. Biotechnol. Bioeng. 24, 1341–1370.
Rittmann, B.E., McCarty, P.L., 1980. Evaluation of steady-state
biofilm kinetics. Biotechnol. Bioeng. 22, 2359–2373.
Rittmann, B.E., McCarty, P.L., 1981. Substrate flux into biofilms of
any thickness. Environ. Sci. Technol. 15, 831–849.
Rittmann, B.E., Snoeyink, V.L., 1984. Achieving biologically
stable drinking water. J. Am. Wastewater Assoc., October
106–114.
Rogers, G.L., Klemetson, S.L., 1985. Ammonia removal in selected
aquaculture water reuse biofilters. Aquacult. Eng. 4, 135–154.
Rusten, B., McCoy, M., Proctor, R., Siljudalen, J.G., 1998. The
innovative moving bed biofilm reactor/solids contact reaeration
process for secondary treatment of municipal wastewater. Water
Environ. Res. 70, 1083–1089.
Sandu, S.I., Boardman, G.D., Watten, B.J., Brazil, B.L., 2002.
Factors influencing the nitrification efficiency of fluidized bed
filter with a plastic medium. Aquacult. Eng. 26, 41–60.
Satoh, H., Okabe, S., Norimatsu, N., Watanabe, Y., 2000. Signifi-
cance of substrate C/N ratio on structure and activity of nitrify-
ing biofilms determined by in situ hybridization and use of
microelectrodes. Water Sci. Tech. 41, 317–321.
213
Shnel, N., Barak, Y., Ezer, T., Dafni, Z., Van Rijn, J., 2002. Design
and performance of a zero discharge tilapia recirculating system.
Aquacult. Eng. 26, 191–203.
Smith, C.E., Williams, W.G., 1974. Experimental nitrate toxicity in
rainbow trout and Chinook salmon. Trans. Am. Fish. Soc. 103,
389–390.
Speece, R.E., 1973. Trout metabolism characteristics and the
rational design of nitrification facilities for water reuse in
hatcheries. Trans. Am. Fish. Soc. 2, 323–334.
Srna, R.F., 1975. A modular nitrification filter design based on a
study of the kinetics of nitrification of marine bacteria. Proc.
World Maricult. Soc. 6, 463–478.
Srna, R.F., Baggaley, A., 1975. Kinetic response of perturbed marine
nitrification systems. J. Water Polut. Control Fed. 47, 472–486.
Summerfelt, S.T., Cleasby, J.L., 1993. Hydraulics in fluidized-bed
biological reactors. In: Wang, J.K. (Ed.), Techniques for Modern
Aquaculture. American Society of Agricultural Engineers, Saint
Joseph, Michigan, pp. 439–457.
Twarowska, J.G., Westerman, P.W., Losordo, T.M., 1997. Water
treatment and waste characterization evaluation of an intensive
recirculating fish production system. Aquacult. Eng. 16, 133–147.
Tucker, C.S., 1985. Channel Catfish Culture. Elsevier Science
Publishers, Amsterdam, The Netherlands.
Watten, B.J., Timmons, M.B., Summerfelt, S.T., Vinci, B.J., 1998.
Comparative performance of biofilm reactor types: application
of steady-state biofilm kinetics. In: Libey, G.S., Timmons, M.B.
(Eds.), Proceedings of the Second International Conference on
Recirculating Aquaculture, Roanoke, Virginia, pp. 355–358.
Watten, B.J., Ridge, W.J., Timmons, M.B., 1993. Evaluation of a
Multi-stage Fluidized Bed Biological Filter for Accelerated
Nitrification and Reduced Energy Demand. American Fisheries
Society Annual Meeting, Pittsburgh, Pennsylvania.
Weatherley, L.R., 1982. Application of simple dynamic response
analysis to a recirculating aquaculture system—a preview.
Aquacult. Eng. 1, 93–113.
Weatherley, L.R., 1984. Rate models for a marine biological filter.
Aquacult. Eng. 3, 15–29.
Weatherley, L.R., Hill, R.G., Macmillan, K.S., 1993. Process mod-
eling of an intensive aquaculture system. Aquacult. Eng. 12,
215–230.
Weber, W.J., 1972. Physiochemical Processes for Water Quality
Control. John Wiley and Sons Inc., New York.
Weng, C., Molof, A.H., 1974. Nitrification in the biological fixed
film rotating disk system. J. Water Pollut. Control. Fed. 46,
1674–1685.
Wheaton, F.W., Hockheimber, J.N., Kaiser, G.E., Krones, M.J.,
Libey, G.S., Easter, C.C., 1994. Nitrification filter principles.
In: Timmons, M.B., Losordo, T.M. (Eds.), Aquaculture Water
Reuse Systems: Engineering Design and Management. Devel-
opments in Aquaculture and Fisheries Science, vol. 27. Elsevier
Science, Amsterdam, The Netherlands, pp. 127–171.
Zhu, S., Chen, S., 1999. An experimental study on nitrification
biofilm performance using a series reactor system. Aquacult.
Eng. 20, 245–259.
Zhu, S., Chen, S., 2001. Effects of organic carbon on nitrification
rate in fixed film biofilters. Aquacult. Eng. 25, 1–11.