You are on page 1of 10

Aquacultural Engineering 34 (2006) 214–223

www.elsevier.com/locate/aqua-online

Linear versus Monod representation of ammonia oxidation


rates in oligotrophic recirculating aquaculture systems
Ronald F. Malone a,*, Jon Bergeron b, Chad M. Cristina c
a
Department of Civil and Environmental Engineering, Louisiana State University,
Baton Rouge, LA 70803, USA
b
U.S. Public Health Service, 4060 Wheaton Way Ste. E., Bremerton, WA 98310, USA
c
Department of Civil Engineering, University of South Alabama, Mobile, AL 36688, USA

Abstract

Monod kinetics are widely used to model nitrifying biofilters. However, these kinetics are incapable of representing the
collapse of volumetric TAN conversion rate (VTR) under high organic loadings. Failure to recognize the underlying
heterotrophic interference can lead to calibration issues as a single Monod function is applied across contrasting levels of
carbon loading. This, plus an historic bias towards the analysis of peak carrying capacities leave modelers poorly prepared to
serve the needs of a mariculture industry demanding oligotrophic designs for broodstock maturation and larval/fingerling
production. Consequently, data was generated by a Monte Carlo technique under the assumption of heterotrophic inhibition to
nitrification. The data was used to compare the accuracy of calibration of the Monod relationship using the traditional
Lineweaver–Burke and Eadie–Hofstee calibration methods against direct linear regression for low substrate (mesotrophic/
oligotrophic) regimes. The results indicate that a simple linear relationship with a zero intercept, calibrated on data ranging from
0.1 to 0.5 g-TAN m3, is most suitable for the representation of the mesotrophic/oligotrophic performance of nitrifying biofilters
based on a comparison of SSE for both the Monte Carlo and field data analyzed herein. Additionally, the coefficient of variation
was found to be between 7 and 8% for the parameter t, which is the slope of the linear relationship between total ammonia
nitrogen (TAN) and VTR while the CV for the Monod parameters ranged between 22 and 143% for VTRmax and between 29 and
137% for the apparent half-saturation constant showing the improved stability of the linear model to that of the Monod model.
# 2005 Published by Elsevier B.V.

Keywords: Monod model; Broodstock; Oligotrophic; Modeling; Inhibition; Nitrification kinetics; Linear regression

1. Introduction Consequently, these systems are increasingly being


used to produce larvae, fry and fingerlings for grow out
The use of recirculating systems for the condition- systems (Malone, 2002). Although this increase in
ing of marine broodstock is gaining wide acceptance. utilization appears to have stemmed from the need for
precise control of temperature and light regimes,
* Corresponding author. adoption now is often driven by the desire to precisely
E-mail address: RMalone@lsu.edu (R.F. Malone). control water quality and nutritional conditions in the

0144-8609/$ – see front matter # 2005 Published by Elsevier B.V.


doi:10.1016/j.aquaeng.2005.08.005
R.F. Malone et al. / Aquacultural Engineering 34 (2006) 214–223 215

sensitive, early stages of life. Also the isolated nature alkalinity on the TAN and nitrite-N conversion
of a closed recirculating system provides benefits in capacities of the biofilters. Although approximate in
the control of diseases that have historically deva- nature, the model proved helpful in the sizing of bead
stated the marine industry. filter and fluidized bed combinations as well as in
Most marine hatchery systems require oligotrophic determining pumping and heating capacities needed to
water quality conditions in order to maintain a healthy support the platform activities.
stock. Broodstock systems generally require the total Certainly, our understanding of nitrification
ammonia nitrogen (TAN) concentration to be in the kinetics has advanced to the point that these models
range of 0.1–0.3 g-N m3. Larval systems typically are useful in a pragmatic sense both as analysis and
require TAN concentrations less than 0.1 g-N m3 and communication tools. However, these same applica-
fingerlings typically require a TAN concentration less tions tend to quickly reveal the shortcomings of our
than 0.5 g-N m3 (Malone and Beecher, 2000). current knowledge and engineering practice, since
Substrate utilization can be divided into the small variations in model parameters may have
following three processes (Zhang and Bishop, significant effects on the projected filter size for a
1994). First, the substrate must diffuse from the bulk given application. The proximate nature of the Monod
fluid, through a static boundary layer, to the biofilm relationships that are used to represent the fixed-film
surface. Under oligotrophic water quality conditions, nitrification kinetics reflects the complexities of
low substrate concentrations reduce the concentration transport and enzymatic reactions that take place in
gradient between the bulk fluid and the static boundary biofilters (Bishop, 1997). The core theory is rooted in
layer, thus limiting the areal nitrification capacities of Michaelis–Menton enzyme kinetics, which can be
biofilters and dramatically increasing their expense. In simply represented as:
the second process, substrate must diffuse through the
Vmax S
biofilm to the embedded bacteria. In the oligotrophic Rate ¼ (1)
kM þ S
water quality conditions, ammonia diffuses into
a relatively thin, vertically homogenous biofilm where Vmax is the maximum specific rate of substrate
that is dominated by autotrophs due to the low utilization (g day1), S the substrate concentration
biochemical oxygen demand (BOD) of the culture (g m3) and kM is the half-saturation constant
water (BOD < 5 g m3 Malone and DeLosReyes, (g m3). Eq. (1) is applicable only to conditions of
1997). Eutrophic conditions create diffusion limita- constant enzyme availability. This relationship is often
tions within the biofilm itself. Biofilms under applied to bioreactors exhibiting dynamic bacterial
these water quality conditions are characterized by growth in the Monod formulation:
layered biofilms that are dominated by relatively fast-
Kmax XS
growing heterotrophs due to the higher BOD Rate ¼ Vb (2)
kM þ S
(BOD > 20 g m3, Malone and DeLosReyes, 1997).
Ammonia must diffuse through the heterotrophic layer where Kmax is the biomass normalized maximum rate
of the biofilm to reach the nitrifiers below the surface. constant in (g of S converted) (g biomass)1 day1, X
Finally, once the substrate reaches the bacteria within is the bacterial biomass (g biomass m3) and Vb is the
the biofilm, these bacteria can consume the substrate. reactor volume (m3).
While sizing tables for oligotrophic conditions are The volumetric TAN conversion rate (VTR) has
known to be approximate (Malone and Beecher, been similarly modeled as (Malone et al., 1993):
2000), the use of computer models to simulate the
Rate VTRmax A
nitrification behavior of a variety of biofilters has been VTR ¼ ¼ (3)
Vb kb þ A
met with some success (Golz et al., 1999). A multi-
compartment computer model was recently developed where VTR is the volumetric TAN conversion rate
to facilitate the design of platform-based broodstock (g m3 day1), VTRmax the maximum VTR attainable
and fingerling production systems to be located off the (g m3 day1), A the TAN concentration in the bulk
coast of California. Compound Monod models fluid (g m3) and kb is the apparent half-saturation
reflected the impact of ammonia, oxygen and constant (g m3). Eq. (3) inherently presumes that the
216 R.F. Malone et al. / Aquacultural Engineering 34 (2006) 214–223

bacterial mass (and in turn the enzyme mass) is measurements under high substrate loading conditions
proportional to the substrate under conditions of and relatively few TAN measurements under low
steady loading (on a time frame of days). Additionally, substrate loading conditions. This bias has been
in fixed-film biofilters, it has been found that the half- further exacerbated by the difficulties associated with
saturation constant is not a universal constant. Rather, obtaining accurate TAN values below 0.3 g m3. The
the value is constant only for a specific set of hydraulic combination of these two biases tends to counteract
conductivity, biofilm thickness and interstitial velocity the limitation of the Lineweaver–Burke method and
conditions and can change with filter management yield coefficients that are biased in favor of the mid-
procedures (Malone et al., 1993). range of TAN values. The end result are calibrations
It is perhaps just fortunate that surface area that are reasonably accurate in predicting filter
dependent diffusion reactions are also well repre- performance under regimes of high nitrogen and
sented by Monod formats. The importance of diffusion organic loading, but, weak or simply erroneous, in the
into biofilms has not only been elegantly recognized oligotrophic zone. Despite the limitations of para-
by early kinetic modelers (Harremoes, 1982), but now meter determination, many researchers have shown
has been verified by direct measurements within that, in a pragmatic sense, the functional relationship
biofilms (Zhang and Bishop, 1994; Zhang et al., 1995; found in Eq. (3) can reasonably represent the VTR of
de Beer et al., 1997). Data is readily available to biofilters over a wide range of TAN concentrations
calibrate Eq. (3), since VTR is easily calculated by (Malone et al., 1993; DeLosReyes and Lawson, 1996;
flow and concentration data (Malone and Beecher, Zhu and Chen, 2001; Golz et al., 1999).
2000). Eq. (3) does not reflect the impact of heterotrophic
Eq. (3) is transformed using either the Eadie– bacteria, which is widely recognized to impair both
Hofstee or the Lineweaver–Burke method. Parameter TAN and nitrite conversion as organic levels in the
values for VTRmax and kb are determined through a system increase (Hockenbury and Leslie Grady, 1977;
linear regression analysis. The linearized versions of Harremoes, 1982; Bovendeur et al., 1990; Zhang and
Eq. (3) using the Eadie–Hofstee and Lineweaver– Bishop, 1994). The algorithms of this model projected
Burke methods can be found in Eqs. (4) and (5), maximum TAN conversion capacities well beyond the
respectively, as: point of known filter failure. Several researchers have
found that VTRmax is not a constant, but is dependent
VTR
VTR ¼ kb þ VTRmax (4) on the carbon to nitrogen (C/N) ratio within the culture
A waters. Satoh et al. (2000) used microelectrodes to
1 kb 1 1 determine the effect of C/N on the spatial distributions
¼ þ (5) of ammonia-oxidizing bacteria. They found that as C/
VTR VTRmax A VTRmax
N increased, NH4+ oxidation rates decreased and
Both methods of linearization have their limita- proposed that interspecies competition between
tions. The Eadie–Hofstee method incorporates VTR heterotrophs and nitrifiers may have been the cause.
into both the ordinate and abscissa in the regression Both Ohashi et al. (1996) and FDZ-Polanco et al.
and is thus more susceptible to measurement error (2000) reached similar conclusions when studying
than is the Lineweaver–Burke method, which does not biological aerated filters. Both teams reported a
include VTR in the ordinate. The Lineweaver–Burke decreasing proportion of nitrifiers in the biofilms as
method uses reciprocals of measured values. If the a function of increasing C/N. Zhu and Chen (2001)
number of data points is spread evenly across the range reported a 70% decline in VTR when the C/N ratio
of interest, this method gives inordinate weight to was increased to 1.0 as compared to a pure ammonia
VTR values when A is less than 1 g m3. feed (C/N = 0). Gupta and Gupta (2001) found that the
The historic emphasis of the recirculating com- nitrification rate increased from 0.81 g-N m2 day1
munity has been focused on freshwater grow out with increasing nitrogen loading despite high organic
systems with high substrate regimes as opposed to the loading. However, they found that the rate of
low substrate regimes required for hatcheries. This nitrification decreased at a high load of 3.35 g-
emphasis resulted in databases with numerous TAN N m2 day1 and 32 g-COD m2 day1 and proposed
R.F. Malone et al. / Aquacultural Engineering 34 (2006) 214–223 217

that the reduction was the result of some form of differing technologies based on a single parameter, the
inhibition process. These findings indicate the need for slope of the regression equation.
the use of some form of inhibition kinetics as opposed
to fundamental Michaelis–Menten kinetics when
modeling these systems. 2. Data generation and analysis
The second shortcoming identified with the current
practice is perhaps an historical artifact of the The limitation of Eq. (3) is that a single equation is
development of biofiltration in aquaculture, which calibrated across two dramatically different operating
has focused more on the heavy production end of the regimes. In a mathematical sense, it is argued that
loading regime than on the oligotrophic. Traditionally, biofilter performance is more appropriately repre-
our principle basis for comparison of filtration capacity sented by a type of inhibition kinetics. To this end,
has been maximum ammonia conversion capacity. The Eq. (6) is presented. However, it should be noted that
maximum conversion capacity was viewed as the the purpose of Eq. (6) is explanatory in nature and is
principle characteristic determining the filters value in not being promulgated as the final form of inhibition
the search for a cost effective, recirculating grow out kinetics in aquacultural engineering. It is solely being
strategy for commodity foodfish. The databases, and in used to highlight the error that can be introduced into
turn, our kinetic constants reflect this biased view VTR modeling by fitting a Monod relationship to a
towards heavily loaded regimes. The increased process whose underlying behavior is governed by
emphasis on the production of marine species has inhibition kinetics.
shifted the demand on our biofilters as they are
Vm A
increasingly used for breeding, larval and fingerling VTR ¼ (6)
kI þ A þ cA2
production. Increasingly the relationship between
extremely low substrate levels and conversion capacity where Vm is the analogous to VTRmax (g m3 day1),
determines the filters sizing, not its peak carrying kI the analogous to the apparent half-saturation con-
capacity. Appropriate determination of the half-satura- stant kb (g m3) and c is the constant (m3 g1) used to
tion constant rather than the maximum conversion provide dimensional consistency.
capacity becomes critical. The methodologies used for Fig. 1 contains sample VTR plots comparing
estimating low substrate (<0.5 g-N m3) conversion Eqs. (3) and (6). The principle difference in the two
capacities must be revised and statistical techniques that relationships is Eq. (6) ability to reflect the collapse of
do not slant curve-fitting exercises towards high nitrification capacity that is associated with increased
substrate concentrations should be identified. organic loading as opposed to the Michaelis–Menten
It is proposed herein that linear regression can be model whose VTR increases and approaches a
used, without data transformation, to model oligo- maximum as A approaches infinity. In this mathema-
trophic systems (Chitta, 1993). Although this method tical configuration, A is used as a surrogate for organic
is not ideal from a purely scientific standpoint, it loading in the inhibition term, noting that the C/N
would have several strengths. First, it would serve as loading ratio remains fixed by the feed protein content.
an intermediate step to improve modeling accuracy in If the focus is shifted from the entire graph to the
oligotrophic systems until a consensus can be reached oligotrophic region (0 g m3  TAN  0.5 g m3), as
on the proper form of inhibition-based nitrification shown in the cut out of Fig. 1, we can notice the
kinetics. Second, since this linear relationship would discrepancy between the estimations of inhibition
only be applied for TAN concentrations in the range of kinetics through the choice of either Michaelis–
0.1–0.5 g m3, data collection could be simplified Menten kinetics or a simple linear regression between
through the use of chemical feeds, since heterotrophic A and VTR. Estimates of the required biofilter size
interference would be minimal under such low using these two different kinetic relationships can
loadings. Third, it would improve the engineer’s differ by as much as 25%.
ability to rapidly size biofilters for oligotrophic The first step in this analysis was data generation.
applications. Finally, it could provide a simple means For this purpose, it was assumed that the inhibition
for potential users of a given biofilter to compare kinetics defined by Eq. (6) truly defined the
218 R.F. Malone et al. / Aquacultural Engineering 34 (2006) 214–223

Fig. 1. Sample plots comparing the estimation error in VTR when the estimation method is either a Monod kinetic model or a direct linear
regression. The inhibition curve represents ammonia oxidation within a biofilter for the given inhibition expression. The Monod and linear curves
represent best-fit estimates of the inhibition curve. Parameter values include Vm = 6000 g m3 day1, kI = 5 g m3, c = 1 m3 g1,
VTRmax = 1392 g m3 day1, kA = 1.231 g m3 and t = 1075 day1. It should be noted that the inhibition expression shown above is for
explanatory purposes.

conditional mean VTR. Eq. (6) produces a continuous a value of 0.3. This value was chosen due to its ability
curve. However, it was necessary to introduce to create reasonable variability in the data while
variability into Eq. (6) to simulate field measurements minimizing the occurrence of negative values for
as: VTR, which was problematic with higher values of f.
  The data were generated in the following manner.
Vm A Vm A N = 20 data sets were generated for each Vm–kI
VTR ¼ N ; f (7)
kI þ A þ cA2 kI þ A þ cA2 combination including values of Vm equal to 6000 and
9000 g m3 day1 and kI = 5 and 10 g m3. Four
where N(Xbar, s) indicates a normal probability density values of VTR were randomly generated from Eq. (7)
function with conditional mean Xbar and standard for each value of A including A = 0.1 to 5.0 g-
deviation s. The constant, c, was defined as TAN m3 in increments of 0.1 g-TAN m3. Four
1 m3 g1. The constant f ( f > 0) represents the stan- values of VTR were generated to more closely
dard deviation as a fixed proportion of the conditional simulate field data where repeated values of A are not
mean. For the purposes of this exploration, f was set to uncommon. Furthermore, four values were found to
R.F. Malone et al. / Aquacultural Engineering 34 (2006) 214–223 219

produce a manageable number of data points for required to sustain a steady state biofilm (Rittmann
analysis. Fig. 2 contains sample plots of the Monte and McCarty, 1980). However, the steady state
Carlo data generated for each step of the analysis. requirement for Amin is not met in this application.
Once the data were generated from Eq. (7), two Fixed-film biofilters are steady state on the order of
methods of estimating the underlying function from days but unsteady state on the order of hours due to
the generated data were attempted. The Lineweaver– feeding. Consequently, values lower than Amin can be
Burke (LB) method in Eq. (5) was used to fit Eq. (3) to sustained for limited durations between feedings, a
the generated data. The equation used to fit the direct finding consistent with those of Furamai and Rittman
linear regression model was expressed as: (1994).
The values of b in Eq. (8) were forced to zero and
VTR ¼ mA þ b (8)
new slopes, t, were computed. The new slopes were
The parameter m is the slope of the line relating used for the remainder of the analysis with the linear
VTR and A while the parameter b is the y-intercept of relationship between VTR and A expressed as:
the equation.
VTR ¼ tA (9)
Least squares analyses were performed to deter-
mine the values of the parameters m and b. When the where t is the normalized conversion capacity relating
20 data sets were analyzed, it was found that b did not VTR and A (day1). Once the value of t has been
statistically differ from 0 for any data set (P > 0.05). determined, the mass of TAN that a biofilter could
This finding was not surprising since the function used convert in one day could be determined.
to generate the data had the property that VTR = 0
Rate ¼ tAVb (10)
when A = 0.
This approach appears to contradict the concept of Under the typical conditions of application where
Amin, the minimum, non-zero, concentration of A both the TAN loading (L) (g day1) and the critical

Fig. 2. Sample plots for the Monte Carlo data generated in this analysis. Each set of data points represents only one out of twenty data sets
generated for each set of Vm and kI. In plot (A) Vm = 6000 g m3 day1, kI = 5 g m3. In plot (B) Vm = 9000 g m3 day1, kI = 5 g m3. In plot
(C) Vm = 6000 g m3 day1, kI = 10 g m3. Finally, in plot (D) Vm = 9000 g m3 day1, kI = 10 g m3.
220 R.F. Malone et al. / Aquacultural Engineering 34 (2006) 214–223

Table 1
Means and coefficient of variations (ratio of standard deviation to mean) for all estimated parameter values
Vm kI Mean estimated Mean estimated Mean estimated
(g m3 day1) (g m3) VTRmax (g m3 day1) (CV%) kA (g m3) (CV%) t (day1) (CV%)
6000 5 1392 (22) 1 (48) 1075 (7)
6000 10 1269 (20) 2 (29) 561 (8)
9000 5 2745 (143) 1 (137) 1666 (8)
9000 10 3336 (137) 4 (121) 870 (7)
N = 20 Monte Carlo data sets were used for each Vm–kI pair to determine mean parameter values.

TAN concentration (AC) (g m3) are user defined, the modeling environment. However, the relatively low
filter size can be directly computed. uncertainty for the DLR leads to greater stability in the
same situation.
Lð1  Is Þ Once the parameter values were determined for a
Vb ¼ (11)
tAC given data set, these parameters were used to
‘‘predict’’ the VTR within the 0.1–0.5 g m3 range.
where Is is the in situ nitrification fraction (unit less) The sum of squared errors (SSE) was computed for
representing that portion of the nitrification that occurs each data set and each kinetic model using only data in
on the tank and pipe walls (Malone and Beecher, this same range. Paired t-tests were then used to
2000). Observed values of Is rarely fall below 30%, determine if either method had a significantly lower
however, this factor is sometimes neglected forcing a SSE than the other for a given Vm–kI combination. It
an oversizing of the filter that contributes to the safety was found that the DLR had a significantly lower SSE
factor of the design. than the Monod model for each of the four data sets
Parameter values for both the Monod and direct (P < 0.05) verifying what was found in Table 1.
linear regression (DLR) methods were estimated for
each of the N = 20 data sets and for each combination
of Vm and kI yielding a total of 80 parameter estimates.
There was one significant difference in this operation. 3. Comparison of methods using field data
The Monod parameters were determined by lineariz-
ing the data across the range of 0.1–5.0 g m3 Field data was sought to verify the results of the
(representative of typical practice) while the DLR Monte Carlo analysis. The data sets included two bead
used only data in the 0.1–0.5 g m3 range (as filter data sets. The first set of data was collected by
proposed). Chitta (1993), while the second data set was collected
The mean parameters for each method and each by Pfeiffer and Malone (2005). The data used in this
Vm–kI pair can be found in Table 1. The coefficient of analysis can be found in Fig. 3 with the results of both
variation shows that parameter estimates for VTRmax the DLR and Monod fitting methods found in Fig. 4.
and kA were more variable than those of t. The Since data was not available values of TAN in excess
standard deviation for t was found to range between 7 of 1.5 g-TAN m3 for the Chitta (1993) data or above
and 8% of the mean value of t. However, the standard 0.8 g-TAN m3 for the Pfeiffer and Malone (2005)
deviation for VTRmax ranged from 22 to 143% of the data, the Monod model was calibrated across the full
mean VTRmax and the standard deviation of kA ranged range of available TAN data for each of the two
from 29 to 137% of the mean kA. To further exacerbate available data sets. The DLR was calibrated only for
the problem, the Monod kinetic expression of Eq. (3) TAN values ranging between 0.0 and 0.5 g-TAN m3,
contains two parameters, both of which are more consistent with the Monte Carlo analysis. The ability
variable than the single-parameter linear model from for the model to mimic the data was tested, using SSE,
Eq. (9). The higher variability in the Monod kinetic only for TAN values ranging between 0.0 and 0.5 g-
model creates uncertainty as to system performance TAN m3, which were again consistent with the
when these systems are modeled in the computer- Monte Carlo analysis.
R.F. Malone et al. / Aquacultural Engineering 34 (2006) 214–223 221

Fig. 3. Field data used to verify Monte Carlo analysis.

The SSE for Chitta (1993) was calculated to be brated to localized data (TAN < 0.5 g m3) was found
474,303 g2-N m6 day2 for the DLR and 508,660 g2- to provide significantly more accurate estimates of
N m6 day2 for the Monod kinetics for values of VTR than Monod kinetics calibrated to the full data
TAN ranging from 0.0 to 0.5 g-TAN m3 The data range (TAN < 5.0 g m3) based on an analysis of the
from Pfeiffer and Malone (2005) resulted in SSEs for sum of squared errors in a Monte Carlo analysis
the DLR and Monod of 3981 and 4280 g2-N m6 (P < 0.05). A similar analysis on field data also
day2, respectively, for the same range of analysis. For indicated that the DLR yielded lower values of SSE
both data sets, the SSE of the DLR was found to be when estimating VTR for TAN values ranging from
lower than that of Monod kinetics. Clearly, this is not 0.0 to 0.5 g-TAN m3. This observation should be
an exhaustive analysis, though it does show the verified with analysis of additional field data from
potential of using a DLR to simplify modeling efforts oligotrophic/mesotrophic systems.
for oligotrophic water quality conditions. The estimated parameter values of the DLR were
found to be less variable than those of the Monod
kinetics based on the coefficients of variation. These
4. Recommendations/suggestions findings indicate that the use of the DLR may
introduce less error into the modeling process than
For the purposes of modeling VTR under does Monod kinetics. Consequently, the proposed
mesotrophic/oligotrophic water quality conditions, premise seems to be valid. The use of the DLR instead
where the TAN concentration must be kept below of Monod kinetics, under oligotrophic conditions,
0.5 g m3, the direct linear regression model cali- could serve as an intermediate step to improve the

Fig. 4. Monod and direct linear regression fits of field data in the oligotrophic water quality range.
222 R.F. Malone et al. / Aquacultural Engineering 34 (2006) 214–223

accuracy of computer models used under these Research Institute of San Diego California. This
conditions until a consensus can be reached on the effort was under the direction of Mark Drawbridge.
proper form of inhibition kinetics. Support for the analysis presented here was provided
The use of t as a rating system would also be by Bioclarifier Engineering Analysis and Design, LLC
valuable to engineers and consumers who are of Baton Rouge Louisiana.
designing or operating oligotrophic systems. These
filters could be compared by their t values from
Table 1. If we only look at the first two Vm–kI entries in
Table 1, we can refer to the first entry as having a t- References
rating of 1075 day1, based on the mean t found in
Bishop, P.L., 1997. Biofilm structure and kinetics. Water Sci.
column 5, while the second entry would be referred to Technol. 36 (1), 27–294.
as having a t-rating of only 561 day1. Immediately, it Bovendeur, J., Zwaga, A.B., Lobee, B.G.J., Blom, J.H., 1990. Fixed-
would be known that the filter rated at 1075 day1 biofilm reactors in aquacultural water recycle systems: effect of
filter was more efficient on a media-volume basis than organic matter elimination on nitrification kinetics. Water Res.
the filter rated at 561 day1 and would require a filter 24 (2), 207–213.
Chitta, B.S., 1993. Effects of backwash frequency on the nitrifica-
comprising only half the volume of the lower rated tion in plastic bead media biofilters used in recirculating finfish
filter. culture systems. M.S. Thesis. Louisiana State University, Baton
It is clear that new forms of nitrification kinetics need Rouge, LA.
to be investigated in order to represent filter collapse de Beer, D., Stoodley, P., Lewandowski, Z., 1997. Measurement of
local diffusion coefficients in biofilms by microinjection and
under high organic loadings. These new kinetics will
confocal microscopy. Biotech. Bioeng. 53, 151–158.
likely resemble the inhibition kinetics presented herein DeLosReyes Jr., A.A., Lawson, T.B., 1996. Combination of a bead
by including some non-linear term within the expres- filter and rotating biological contactor in a recirculating fish
sion. However, it is unclear if fitting an equation across a culture system. Aquacult. Eng. 15 (1), 27–39.
broad range of TAN loadings will ever prove more FDZ-Polanco, F., Mendez, E., Uruena, M.A., Villaverde, S., Garcia,
accurate than fitting a linear model across the narrow P.A., 2000. Spatial distribution of heterotrophs and nitrifiers in a
submerged biofilter for nitrification. Water Res. 34 (16), 4081–
oligotrophic/mesotrophic range. Since the focus of this 4089.
paper is on mesotrophic/oligotrophic systems, the Furamai, H., Rittman, B.E., 1994. Interpretation of bacterial activ-
parameter, t, was calculated for TAN concentrations ities in nitrification filters by a biofilm model considering the
ranging between 0.1 and 0.5 g-TAN m3. However, it kinetics of soluble microbial products. Water Sci. Technol. 30
may be possible to extend the analysis beyond the range (11), 147–156.
Golz, W.J., Rusch, K.A., Malone, R.F., 1999. Modeling the major
that was analyzed in this study. limitations on nitrification in floating-bead filters. Aquacult.
In summary, it is the author’s opinion that the linear Eng. 20, 43–61.
representation of data generated in the oligotrophic Gupta, A.B., Gupta, S.K., 2001. Simultaneous carbon and nitrogen
and mesotrophic range will provide a more accurate removal from high strength domestic wastewater in an aerobic
RBC biofilm. Water Res. 35 (7), 1714–1722.
representation than the broadly calibrated Monod
Harremoes, P., 1982. Criteria for nitrification in fixed film reactors.
kinetics. The development of a form of Monod Water Sci. Technol. 14, 167–187.
kinetics that is capable of representing heterotrophic Hockenbury, M.R., Leslie Grady Jr., C.P., 1977. Inhibition of
interference under eutrophic conditions is deemed nitrification-effects of selected organic compounds. J. Water
appropriate. Additional focus on data collection in the Pollut. Control Fed. 49, 768–777.
oligotrophic zone is warranted. These data should be Malone, R.F., DeLosReyes, A.A., 1997. Categories of recirculating
aquaculture systems. In: Timmons, M.B., Losordo, T. (Eds.),
used to confirm the assumption of zero intercept and to Advances in Aquacultural Engineering: Proceedings from the
verify the observations presented here. Aquacultural Engineering Society (AES) Technical Sessions at
the Fourth International Symposium on Tilapia in Aquaculture,
Coronado Springs Resort, Walt Disney World, Orlando, Fl,
Acknowledgements November 9–12, 1997, pp. 197–208.
Malone, R.F., Chitta, B.S., Drennan, D.G., 1993. Optimizing nitri-
fication in bead filters for warmwater recirculating aquaculture
This work was stimulated by a modeling effort systems. In: Wang, J.K. (Ed.), Techniques in Modern Aqua-
conducted under funding by Hubbs-SeaWorld culture (Proceedings), 21–23, June, Spokane, WA American
R.F. Malone et al. / Aquacultural Engineering 34 (2006) 214–223 223

Society of Agricultural Engineers, St. Joseph, MI, pp. 315– biofilter in a recirculating warmwater fish system. Aquacult.
325. Eng. 34 (3), 311–321.
Malone, R.F., Beecher, L.E., 2000. Use of floating bead filters to Rittmann, B.E., McCarty, P.L., 1980. A model of steady-state-
recondition recirculating waters in warmwater aquaculture pro- biofilm kinetics. Biotech. Bioeng. 22, 2343–2357.
duction systems. Aquacult. Eng. 22, 57–74. Satoh, H., Okabe, S., Norimatsu, N., Watanabe, Y., 2000. Signifi-
Malone, R.F., 2002. Perspectives on responsible aquaculture for the cance of substrate C/N ratio on structure and activity of nitrify-
new millennium. In: Creswell, R.L., Flos, R. (Eds.), World ing biofilms determined by in situ hybridization and the use of
Aquaculture Society, Baton Rouge. LA USA and European microelectrodes. Water Sci. Technol. 41 (4), 317–321.
Aquaculture Society, Oostende, Belgium, pp. 94–111. Zhang, T.C., Bishop, P.L., 1994. Experimental determination of the
Ohashi, A., Viraj de Silva, D.G., Mobarry, B., Manem, J.A., dissolved oxygen boundary layer and mass transfer resistance
Stahl, D.A., Rittmann, B.E., 1996. Influence of substrate C/ near the fluid-biofilm interface. Water Sci. Technol. 30 (11), 47–
N ratio on the structure of multi-species biofilms consisting 58.
of nitrifiers and heterotrophs. Water Sci. Technol. 32 (8), Zhang, T.C., Fu, Y.C., Bishop, P.L., 1995. Competition for substrate
75–84. and space in biofilms. Water Environ. Res. 67, 992–1003.
Pfeiffer, T.J., Malone, R.F., 2005. Nitrification performance of a Zhu, S., Chen, S., 2001. Effects of organic carbon on nitrification
propeller-washed bead clarifier supporting a fluidized sand rate in fixed film biofilters. Aquacult. Eng. 25, 1–11.

You might also like