See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/265434013

Simulation of Monods model with Mathematica

7.0 (Fizmako research group)
Article January 2010

READS

29

2 authors:
Alvaro H. Salas

Leonel L. Palom P.

National University of Colombia

National University of Colombia

124 PUBLICATIONS 856 CITATIONS

1 PUBLICATION 0 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate,

letting you access and read them immediately.

SEE PROFILE

Available from: Alvaro H. Salas

Retrieved on: 19 July 2016

Far East Journal of Applied Mathematics

Volume 47, Number 2, 2010, Pages 125-135
This paper is available online at http://pphmj.com/journals/fjam.htm
2010 Pushpa Publishing House

SIMULATION OF MONODS MODEL WITH

MATHEMATICA 7.0
(FIZMAKO RESEARCH GROUP)
ALVARO H. SALAS S.1,2 and LEONEL L. PALOM P.1,2
1Departamento

de Matemticas
Universidad de Caldas
Colombia
e-mail: asalash2002@yahoo.com

2Departamento

de Matemticas y Estadstica
Universidad Nacional de Colombia-sede Manizales
Colombia
e-mail: llpalomap@unal.edu.co
Abstract
The Monods model for microbial growth is studied analytically. Starting
from this analysis we simulate it with the aid of Mathematica 7.0.
Simulations are presented graphically and numerically. A Mathematica
7.0 program is included in Appendix.

1. Introduction
As it was remarked (Strigul et al. [25]) the Monod model was suggested by
Nobel Laureate J. Monod in 1942 and for more than 60 years has been one of the
most frequently used models in microbiology (Monod [17]; Pirt [19]; Koch [14];
Kovarova-Kovar and Egli [15]). Most models of chemostat growth are based on the
Monod equations (Pirt [19]; Smith and Waltman [23]), and numerous models of
2010 Mathematics Sub ject Classification: 92C40.

Keywords and phrases: Monods model, microbial ecology, microbial growth, Mathematica
7.0.
Received July 24, 2010

126

ALVARO H. SALAS S. and LEONEL L. PALOM P.

microbial ecology incorporate Monods growth kinetics (Koch [14]; Strigul and
Kravchenko [24]). One of the very important practical applications of this model is
the evaluation of the biodegradation kinetics of organic pollutants in environmental
systems (Blok [2]; Blok and Struys [3]). The Monod model describes microbial
growth with three parameters: (1) maximal specific growth rate; (2) a saturation
constant; (3) a yield coefficient. In the case of biodegradation kinetics, these
parameters can be used as criteria for the biodegradability of organic pollutants
(Blok [2]).
One of the most important problems in practical applications of the Monod
model is evaluating the parameters values from experimental data. It was shown
that in many cases, reasonable estimates of the parameters cannot be obtained by a
simple application of non-linear least squares estimators (Holmberg [10]; Saez and
Rittmann [21]). A possible way to circumvent this problem is the application of
optimal experimental design techniques (Dette et al. [6, 7]). Optimal design theory is
a branch of statistics actively developed since 1960, when basic results by Kiefer
and Wolfowitz were published (Kiefer and Wolfowitz [12]; Kiefer [13]). In general,
the application of optimal experimental designs can be useful in (1) reducing the
number of necessary experimental measurements; (2) improving the precision of
model parameter value determinations (Fedorov [8]). Optimal design theory is well
developed for linear regression models, and for several simple non-linear regression
models (Dette et al. [6, 7]). Recently, a complete theoretical examination of optimal
designs for the Monod model was presented (Dette et al. [5, 6]), where two different
approaches to the problem of designing experiments for the Monod model were
investigated, namely local optimal experimental designs (Dette et al. [5]); and
maximin optimal experimental designs (Dette et al. [6]).
2. The Monod Model for a Simple Batch Culture and its Properties
In a simple homogeneous batch culture it is assumed that the growth conditions
are similar for all cells (Pirt [19]). Traditionally, a typical growth curve is divided
into six phases: (1) lag, (2) accelerating, (3) exponential, (4) decelerating, (5)
stationary, and (6) declining growth (Monod [17]). In many cases, cultures do not
demonstrate this typical growth; also, the experimenter might be interested in only
particular aspects of growth: for instance, in predictive microbiology the duration of
the lag phase is the main parameter characterizing growth inhibition. This has led to
models simulating only a few focal phases of growth, for example in predictive
microbiology (Dette et al. [7]), and to more sophisticated structured models able to

SIMULATION OF MONODS MODEL WITH MATHEMATICA 7.0

127

describe typical growth curves in general (Koch [14]). The role of the experimenter
is to select the model that best reflects the characteristics of microbial growth that
are his current focus, and to balance a models complexity and its flexibility.
The Monod model uses a very convenient approximation of the batch growth
process, especially when describing biodegradation kinetics (Blok [2]; Dette et al.
[7]); it recognizes only 3 growth phases, i.e., exponential and decelerating growth,
and a stationary phase. It is assumed that lag and accelerating and declining growth
phases do not exist. This assumption is satisfied in many practical cases. The cells
used to inoculate batch cultures are often taken from another actively growing
culture, and, therefore, growth starts immediately in the exponential phase. However,
one should not try to obtain Monods model parameters directly from experimental
data if a lag phase is observed. In this case, it is important to estimate the duration of
the lag phase in preliminary experiments and take the time when this phase ends as a
starting moment of the Monod experiment. The declining growth phase is usually
not a consideration in biodegradation studies, but it can be analyzed separately,
using, for example, the negative exponential model, for which optimal experimental
designs have been also constructed (Dette et al. [7]). Another important assumption
of the Monod model is that microbial growth is limited solely by the substrate
concentration and, therefore, it is very important that experimental conditions satisfy
this requirement. The Monod model suggests that the microbial growth rate (t ) and
the substrate concentration s(t ) at time t are related by the Michaelis-Menten
function
(t ) =

u max s(t )
.
k s + s(t )

(1)

The microbial growth rate for the population x(t ) is then given by the equation
x(t ) = (t ) x(t ).

(2)

Finally, the model assumes that some constant fraction of the consumed substrate is
transformed into microbial biomass
s(t ) =

1
x(t ).
Yx s

(3)

In these equations the three model parameters are: maximal specific growth rate

(u max ), saturation constant (k s ) and yield coefficient (Yx s ).

128

ALVARO H. SALAS S. and LEONEL L. PALOM P.

The Monod differential equations (1)-(3) can easily be integrated. Indeed, from

(3),
s(t ) = px(t ) + C , where p =

1
and C = const.
Yx s

(4)

Inserting equations (1) and (4) into equation (2) gives

x(t ) =

dx u max ( px(t ) + C )
x(t ).
=
dt
k s + px(t ) + C

(5)

Separating variables in (5) it results

dt =

k s + px + C
dx, x = x(t ).
u max ( px + C ) x

(6)

Integrating this equation we obtain

C
C ln x k s ln p +
x

t =
+ D, where D = const.
Cu max

(7)

Equation (7) defines x as a function of t implicitly. The constants C and D may be

determined from the initial conditions x0 = x(0) (the initial microbial population)
and s0 = s(0 ) (the initial substrate concentration). Let t = 0 in (4). Then
C = s0 px0 .

(8)

On the other hand, the constant D is found by letting t = 0 in equation (7) and
taking into account equations (4) and (8),
s
C
k s ln p +
k s ln 0 ( s0 px0 ) ln x0
C ln x0
x0
x0

=
D=
( s0 px0 ) k s
Cu max

or
D=

k s ln s0 ( s0 px0 + k s ) ln x0
.
( s0 px0 ) k s

(9)

Substituting (8) and (9) into (7) and simplifying yields

t = t(x) =

1
u max ( x0 + s0Yx s )

s0Yx s
x

x
( x0 + s0Yx s ) ln + k sYx s ln
,
x
x
x
+
s
Y

x
0
0 x s
0 0

(10)

SIMULATION OF MONODS MODEL WITH MATHEMATICA 7.0

129

0 < x < x0 + s0Yx s .

(11)

t ( x ) = + ,

(12)

where

Observe that
lim

x x0 + s0Yx s

lim t ( x ) =

(13)

x 0+

and
x ( x0 + Yx s ( s0 + k s ))
dt
=
> 0 for all x (0, x0 + s0Yx s ).
dx u max x( x ( x0 + s0Yx s ))

(14)

From (12), (13) and (14), we conclude that function defined by equation (10) is
strictly increasing, so there exists its inverse function x = x(t ) for < t < . But
t 0 in the case we are considering. Since x(0 ) = x0 , the range of the function
x = x(t ) for t 0 is the interval [ x0 , x0 + s0Yx s ).

On the other hand, in view of (12) the line x = x0 + s0Yx

is a vertical

asymptote to the curve described by equation (10) and then this line is a horizontal
asymptote to the curve described by the equation x = x(t ). Now, we proceed to
analyze the concavity of this curve. We have
2
( x x0 s0Yx s ) ( x 2 2((k s + s0 ) Yx s + x0 ) x
xumax

x(t ) =

+ ( s0Yx s + x0 ) ((k s + s0 ) Yx s + x0 ))

( x x0 (k s + s0 ) Yx s )3

(15)

For an inflection point we must have x(t ) = 0. Solving this equation gives
x1, 2 = x0 + s0Yx s + k sYx s k sYx s ( x0 + s0Yx s + k sYx s ).

(16)

We have
x1 = x0 + s0Yx s + k sYx s + k sYx s ( x0 + s0Yx s + k sYx s ) > x0 + s0Yx s , (17)

ALVARO H. SALAS S. and LEONEL L. PALOM P.

130

so x1 (0, x0 + s0Yx s ). On the other hand,

x2 = x0 + s0Yx s + k sYx s k sYx s ( x0 + s0Yx s + k sYx s )
=

(18)

( x0 + s0Yx s ) ( x0 + s0Yx s + k sYx s )

(0, x0 + s0Yx s ). (19)
x0 + s0Yx s + k sYx s + k sYx s ( x0 + s0Yx s + k sYx s )

Thus, the curve x = x(t ) has an inflection point at t = t2 , where t 2 = t ( x2 ).

This analysis allows us to simulate the model. This is our next purpose.
3. Simulation of the Monods Model

Function defined by equation (10) has remarkable properties (Pirt [19]):

1. It is a sigmoidal growth curve with one inflection point.
2. The microbial biomass tends to a horizontal asymptote x = x0 + s0Yx s . This
value is the microbial biomass at the stationary state. Only the yield coefficient
determines the horizontal asymptote. The other two parameters do not affect it. It is
necessary to emphasize two important assumptions that should be taken into
consideration by a practitioner attempting to identify Monods parameter values
from experimental data. First, the sigmoidal function, which is a solution of the
Monod model, starts in the exponential growth phase. Therefore, an experimenter
must exclude measurements reflecting a lag phase. It is equivalent to shifting the
experiment starting time to the right, to a time equal to the observed lag phase.
Second, a solution of the Monod model demonstrates an infinite stationary phase,
where microbial biomass tends to approach a stationary value. However, this is not
observed in real life. Typically, a stationary phase is not observed for a long time,
and, sometimes, it occurs only for a very short time, followed by a decline in
microbial mass. Nevertheless, it is very important to identify the microbial biomass
(or microbial concentration) at the stationary phase, since this value determines the
yield coefficient. Therefore, when determining the Monod parameter values, a
practitioner should determine when the stationary phase starts and the biomass at
that moment, ignoring any decline in biomass that may follow.
In order to simulate the model, we graph the obtained solution (10) for the
following values of the parameters:

SIMULATION OF MONODS MODEL WITH MATHEMATICA 7.0

x0 = 0.03 mg l, s0 = 1 mg l,

1
u max = 0.25 h , k s = 0.5 mg l, Yx

= 0.25 mg mg.

131

(20)

These values give (see equations (4), (8) and (9))

p = 4,

C = 1.12,

D = 20.2879

(21)

and then equation (7) reduces to

1.12
1.12 ln x 0.5 ln 4 +

t =
+ 20.2879, 0.03 x < 0.28.
0.28

(22)

We have an inflection point at (12.0028, 0.18). This point may be calculated by

formula (18). The respective graph is shown below. Observe that the line x = 0.28
is a horizontal asymptote to the curve.

Apart from the graphical information, it is convenient to get numerical data.

This may be done by tabulating the solution. In the table below we present this
useful information.

132

ALVARO H. SALAS S. and LEONEL L. PALOM P.

4. Conclusions

We successfully analyzed Monods model in detail. By using Real Analysis

techniques, we indicated the domain and the range of the solutions to differential
equations. With the aid of Mathematica 7.0, we simulated the model both graphically
and numerically. Due to the importance of Monods model, we think that the results
presented in this paper may be useful in the study of microbial growth.

SIMULATION OF MONODS MODEL WITH MATHEMATICA 7.0

133

Appendix: Mathematica Program for Monods Model

Usage: If we need to simulate the model for

x0 = 0.03 mg l, s0 = 1 mg l,

1
u max = 0.25 h , k s = 0.5 mg l, Yx s = 0.25 mg mg,
we write monod[0.5, 0.25, 0.25, 0.03, 1, 0.01] in Mathematica. This will give the
graph of the solution and a table with numerical information. The last parameter
paso = 0.01 is the step. For example, in the resulting table the first value of x(t )

will be 0.03, the second one will be 0.03 + 0.01 = 0.04 and so on. Of course, the
smallest the value of paso, the largest the table.
References
[1]

R. Berkholz, D. Rohlig and R. Guthke, Data and knowledge based experimental

design for fermentation process optimization, Enzyme Microb. Tech. 27(10) (2000),
784-788.

ALVARO H. SALAS S. and LEONEL L. PALOM P.

134
[2]

J. Blok, Classification of biodegradability by growth kinetic parameters, Ecotoxicol.

Environ. Saf. 27 (1994), 294-305.

[3]

J. Blok and J. Struys, Measurement and validation of kinetic parameter values for
prediction of biodegradation rates in sewage treatment, Ecotoxicol. Environ. Saf. 33(3)
(1996), 217-227.

[4]

S. C. Chapra, Surface Water Quality Modeling, McGraw-Hill Science, New York,

1997.

[5]

H. Dette, V. B. Melas, A. Pepelyshev and N. Strigul, Efficient design of experiments

in the Monod model, J. R. Stat. Soc. B 65 (Part 3) (2003), 725-742.

[6]

H. Dette, V. B. Melas, A. Pepelyshev and N. Strigul, Design of experiments in the

Monod model - robust and efficient designs, J. Theor. Biol. 234(4) (2005a), 537-550.

[7]

H. Dette, V. B. Melas and N. Strigul, Application of optimal experimental design in

microbiology, Optimal Designs: Their Roles and Applications, W. Wong and M.
Berger, eds., Wiley and Sons Ltd., New York, 2005b, pp. 137-180.

[8]
[9]

V. V. Fedorov, Theory of Optimal Experiments, Academic Press, 1972.

M. E. Hidalgo and E. Ayesa, A numerical identifiability test for state-space models application to optimal experimental design, Water Sci. Tech. 43(7) (2001), 339-346.

[10]

A. Holmberg, On the practical identifiability of microbial growth models

incorporating Michaelis-Menten type nonlinearities, Math. Biosci. 62(1) (1982),
23-43.

[11]

R. I. Jenrich, Asymptotic properties of nonlinear least squares estimators, Ann. Math.

Statist. 40 (1969), 633-643.

[12]

J. Kiefer and J. Wolfowitz, The equivalence of two extremum problems, Can. J. Math.
14 (1960), 363-366.

[13]

J. C. Kiefer, General equivalence theory for optimum designs (approximate theory),

Ann. Stat. 2 (1974), 849-879.

[14]

A. L. Koch, The Monod model and its alternatives, Mathematical Modeling in

Microbial Ecology, A. L. Koch, J. Robinson and G. A. Milliken, eds., Chapman and
Hall, New York, 1997, pp. 62-93.

[15]

K. Kovarova-Kovar and T. Egli, Growth kinetics of suspended microbial cells: from

single-substrate-controlled growth to mixed-substrate kinetics, Microbiol. Mol. Biol.
Rev. 62(3) (1998), 646-666.

[16]

W. Merkel, A. Schwarz, S. Fritz, M. Reuss and K. Krauth, New strategies for

estimating kinetic parameters in anaerobic wastewater treatment plants, Water Sci.
Tech. 34(5-6) (1996), 393-401.

SIMULATION OF MONODS MODEL WITH MATHEMATICA 7.0

135

[17]

J. Monod, The growth of bacterial cultures, Ann. Rev. Microbiol. 3 (1949), 371-393.

[18]

A. Munack, Optimization of sampling, Biotechnology, Vol. 4, VCH, Weinheim, 1991,

pp. 252-264.

[19]

S. J. Pirt, Principles of Microbe and Cell Cultivation, Wiley, New York, 1975.

[20]

W. H. Press, S. A. Teukolsky, W. T. Vetterling and B. P. Flannery, Numerical

Recipes: The Art of Scientific Computing, Cambridge University Press, New York,
2007.

[21]

P. B. Saez and B. E. Rittmann, Model-parameter estimation using least squares, Water

Res. 26(6) (1992), 789-796.

[22]

I. Y. M. Smets, K. J. E. Versyck and J. F. M. Van Impe, Optimal control theory: a

generic tool for identification and control of (bio-)chemical reactors, Ann. Rev.
Control 26(1) (2002), 57-73.

[23]

H. L. Smith and P. Waltman, The Theory of the Chemostat: Dynamics of Microbial

Competition, Cambridge University Press, 1995.

[24]

N. S. Strigul and L. V. Kravchenko, Mathematical modeling of PGPR inoculation into

the rhizosphere, Eviron. Model. Softw. 21(8) (2006), 1158-1171.

[25]

N. Strigul, H. Dette and V. B. Melas, A practical guide for optimal designs of

experiments in the Monod model, Environ. Model. Softw. 24(9) (2009), 1019-1026.

[26]

N. S. Strigul, A. Koutsospyros, P. Arienti, C. Christodoulatos, D. Dermatas and W.

Braida, Effects of tungsten on environmental systems, Chemosphere 61(2) (2005),
248-258.

[27]

P. A. Vanrolleghem, M. Van Daele and D. Dochain, Practical identifiability of a

biokinetic model of activated sludge respiration, Water Res. 29(11) (1995),
2561-2570.

[28]

K. J. Versyck, K. Bernaerts, A. H. Geeraerd and J. F. Van Impe, Introducing optimal

experimental design in predictive modeling: a motivating example, Int. J. Food
Microbiol. 51(1) (1999), 39-51.