Res. Lett. Inf. Math.

Sci, (2001) 2, 23-46

Available online at http://www.massey.ac.nz/wwiims/~rlims

Analysis of Logistic Growth Models

A.Tsoularis,
I.I.M.S., Massey University Albany Campus, Auckland, New Zealand
[email protected]
Abstract
A variety of growth curves have been developed to model both unpredated, intraspecific
population dynamics and more general biological growth. Most successful predictive models are
shown to be based on extended forms of the classical Verhulst logistic growth equation. We further
review and compare several such models and calculate and investigate properties of interest for
these. We also identify and detail several previously unreported associated limitations and restrictions.
A generalized form of the logistic growth curve is introduced which is shown incorporate these
models as special cases. The reported limitations of the generic growth model are shown to be addressed
by this new model and similarities between this and the extended growth curves are identified.
Several of its properties are also presented. We furthermore show that additional growth characteristics
are accommodated by this new model, enabling previously unsupported, untypical population dynamics to
be modelled by judicious choice of model parameter values alone.
1. Introduction
In order to model growth of biological systems numerous models have been introduced. These
variously address population dynamics, either modelled discretely or, for large populations, mostly
continuously. Others model actual physical growth of some property of interest for an organism
or organisms.
The simple exponential growth model can provide an adequate approximation to such growth for
the initial period. However, for populations, no predation or intraspecific competition is included.
The population would therefore continue to increase unhindered (or inevitably reduce to zero if
an initial growth reduction were present). Even in the case where predation was at most
negligible, the model does not accommodate reductions due to intraspecific competition for
environmental resources such as food and habitat. For the case of growth per se, unrestricted
growth is also unrealistic. For example, as plants approach maturity, the physical characteristics of
interest will reach a limiting dimension.
Verhulst [1] considered that, for the population model, a stable population would consequently
have a saturation level characteristic: this is typically called the carrying capacity, K, and forms a
numerical upper bound on the growth size. To incorporate this limiting form he introduced the
logistic growth equation which is shown later to provide an extension to the exponential model.
This logistic equation can also be seen to model physical growth provided K is interpreted, rather
naturally, as the limiting physical dimension. It is parameterized by the initial population size (or
physical dimension), the initial growth rate, and K. For typical values of these, particularly, where
the initial population size (or dimension) is smaller than K, the resulting logistic growth rate
curve is sigmoidal. Furthemore, the point of inflection for the system is fixed such that the
corresponding population (or dimension) is

K
. This places an undesirable restriction on the
2

shape of the curve and clearly limits the generality of the model. We will see later that the
Verhulst logistic growth model has formed the basis for several extended models. Each is a
parameterised version of the original and provides a relaxation of this restriction.
Notwithstanding this limitation the logistic growth equation has been used to model many diverse
biological systems. Carlson [2] reported the growth of yeast which is modelled well by the curve
[3,4]. Morgan [5] ingeniously used the equation to describe herding behaviour of African

24

R.L.I.M.S. Vol. 2, May 2001

elephants. Krebs [6] also used the Verhulst logistic equation to fit to population data for
Peruvian anchovies.
There have been applications of the logistic model outside the field of Biology also. Fisher and
Pry [7] have successfully exploited the logistic model to describe the market penetration of many
new products and technologies. In this particular application of the logistic model N represents a
measure of the market already captured and

KN
that for the fraction of the market remaining
K

to be captured. Marchetti and Nakicenovic [8] have given a summary of world energy usage and
source substitution by employing the logistic model. Herman and Montroll [9] have shown that
as basic an evolutionary process as the industrial revolution may also be modelled by logistic
dynamics. Here, as the industrial revolution evolved, the fraction of the labour force in agriculture
declined while the fraction in industry grew.
In this paper we present the major models of this form. We identify the population (or
dimension) corresponding to the characteristic point of inflection for each and compare and contrast
these. We also present previously unreported properties of these models. Finally, we introduce a
generalized logistic equation which incorporates all these as special cases. We also investigate
properties of this model.
2. The Logistic Growth Curve
The simplest realistic model of population dynamics is the one with exponential growth

dN
= rN
dt
with solution

N (t ) = N 0 e rt
where r is the intrinsic growth rate and represents growth rate per capita. To remove unrestricted
growth Verhulst [1] considered that a stable population would have a saturation level
characteristic of the environment. To achieve this the exponential model was augmented by a
multiplicative factor,

N
, which represents the fractional deficiency of the current size from
K

the saturation level, K.

In Lotkas analysis [10] of the logistic growth concept the rate of population growth,

dN
, at
dt

any moment t is a function of the population size at that moment, N(t), namely,

dN
= f (N )
dt
Since a zero population has zero growth, N=0 is an algebraic root of the yet unknown function
f(N). By expanding f(N) as a Taylor series near N=0 and setting f(0)=0 Lotka obtained the
following power series

N2
f ( N ) = Nf (0) +
f (0)
2
N

f (0)
= N f (0) +
2

where higher terms are assumed negligible.

By setting f (0) = r and f (0) =

2r
, where r is the intrinsic growth rate of the population
K

and K is the carrying capacity, one is led to the Verhulst logistic equation

A. Tsoularis, Analysis of Logistic Growth Models

25

dN
N

= rN 1
dt
K

(1)

The Verhulst logistic equation is also referred to in the literature as the Verhulst-Pearl equation after
Verhulst, who first derived the curve, and Pearl [11], who used the curve to approximate
population growth in the United States in 1920.
Equation (1) has solution

N (t ) =

KN 0
( K N 0 )e rt + N 0

(2)

where N0 is the population size at time t = 0.

The three key features of the logistic growth are:
(i)
lim N (t ) = K , the population will ultimately reach its carrying capacity.
t

1 dN
, declines linearly with increasing population size.
N dt

(ii)

The relative growth rate,

(iii)

The population at the inflection point (where growth rate is maximum), Ninf , is exactly
half the carrying capacity, Ninf

K
.
2

For r > 0, the resulting growth curve has a sigmoidal shape and, from (2), is asymptotic to the
carrying capacity. When r < 0 and a reduction in the growth rate per capita is present, the growth curve is
asymptotic to zero leading to population extinction. In the trivial case of no intrinsic growth rate, r = 0, the
population remains static at the initial value of N0. Population biologists and ecologists are interested
mainly in the case where r > 0 and we restrict our investigations to this case in this paper.
Figure 1 depicts several logistic curves for various such values of r with N0=10, K=100. The
larger the r the faster in time the curve reaches the carrying capacity K. Figure 2 illustrates the
fact that the population at the inflection point, Ninf =

K
= 50 , regardless of the value assumed
2

by the intrinsic growth rate, r.

100
r=10
r=2.5

90
r=3
80

r=1

r=1.5

r=0.9

r=5

70

60

50

40

30

20

10

26

R.L.I.M.S. Vol. 2, May 2001

Figure 1. The evolution of population size in time for the Verhulst logistic growth.
250
r=10

200

150

r=5
100

50

r=1
0
10

20

30

40

50

60

70

80

90

100

Figure 2. Plots of the growth rate versus population size for the Verhulst logistic growth.
3. Extended Logistic Growth Models
Since the original work of Verhulst [1] and Pearl and Reed [11] there have been several
contributions suggesting alternative functional forms, f(N), for growth whilst retaining the sigmoid
and asymptotic property of the Verhulst logistic curve. In the plant sciences, Richards [12] was
the first to apply a growth equation developed first by Von Bertalanffy [13] to describe the
growth of animals. Richards growth curve was used for fitting experimental data by Nelder [14],
who used the term generalized logistic equation to describe the equation. Blumberg [15]
introduced the hyperlogistic equation as a generalization of Richards equation. Turner and coauthors [16,17] suggested a further generalization of the logistic growth and termed their
equation the generic logistic equation. In a more recent survey paper Buis [18] revisited the
previous works on logistic growth functions and outlined some of their respective properties.
In this section we derive several well known growth functions which extend the standard Verhulst
equation. In addition, we examine the presence or absence of the sigmoid feature that
characterizes most growth curves and is responsible for the existence of an inflection point.
3.1 Generic Growth Function
Turner and co-authors [17] proposed a modified Verhulst logistic equation which they termed
the generic growth function. This has the form

N
dN
= rN 1+ (1 ) 1
dt
K
where , are positive exponents and

< 1+

1
. This has the solution,

(3)

A. Tsoularis, Analysis of Logistic Growth Models

27

N (t ) =

1 1

1
1 + ( 1) rK (1 ) t +
N 0

The population at the inflection point, Ninf , is given by

N inf

= 1
1+

(4)

For = =1 the functional form for Ninf reduces to that for the Verhulst logistic equation. The
condition

< 1+

ensures that Ninf > 0. For extreme values of and we obtain the

following limits for Ninf :

lim N inf = Ke , 0 < <

0

lim N inf = K , 0 < < 1

lim N inf = K , 0 < <

0

lim N inf = 0 , 0

Figure 3 displays several generic growth curves evolving in time t and figure 4 presents the
growth rate versus time evolution for N0 = 10, K = 100. A visible inflection point occurs for =
0.2

5.0, = 1.0, given by

N inf

1
= K 70 , and is clearly seen in Figure 4. For = 0.5, =
6

2.5,

1
N inf = K 2.5 < 10 , and no inflection is present. Also no inflection occurs for =
6
5

0.2, = 5.0,

1
N inf = K 0 . An inflection point, Ninf = 25, is present for = 1.0, = 1.5.
6

More symmetric graphs with inflection occurring at around half the value of K, are those with
= 2.0, = 1.0 and = 3.0, = 1.0, giving Ninf 58 and Ninf 63 respectively.

28

R.L.I.M.S. Vol. 2, May 2001

3. Generic Growth Function

100

90

beta=1.0
gamma=1.5

beta=5.0
gamma=1.0

80

beta=0.5
gamma=2.5

beta=1.0
gamma=1.7

Population size N

70

60

50

40

30

beta=0.2
gamma=5.0

20

10

20

40

60

80

100

120

Time t

Figure 3. The evolution of population size in time for several parameter pairs ( , ) according
to the generic growth form.

4. Generic Growth
400
beta=2.0
gamma=1.0
350
beta=0.5
gamma=2.5

Growth rate dN/dt

300
beta=3.0
gamma=1.0

250

beta=5.0
gamma=1.0

200

150

100

beta=1.0,gamma=1.5

50

20

40

60

80

100

Population size N

Figure 4. Growth rate versus size plots for the generic growth function.
3.2. Blumbergs equation

120

A. Tsoularis, Analysis of Logistic Growth Models

29

Blumberg [15] introduced another growth equation based on a modification of the Verhulst
logistic growth equation to model population dynamics or organ size evolution.
Blumberg
observed that the major limitation of the logistic curve was the inflexibility of the inflection
point. He further observed that attempts to modify the constant intrinsic growth rate term, r,
treating this as a time-dependent polynomial to overcome this limitation, often leads to
underestimation of future values (see also [10]). Blumberg therefore introduced what he called the
hyperlogistic function, accordingly

N
dN

= rN 1
dt
K

(5)

Blumbergs equation is consistent with the Turner and co-authors generic equation (3) when =
2 - , = 1, and < 2.
Equation (5) can be re-formulated as the integral equation
N (t )
K

(1 x) dx = rK 1t

N0
K

This does not always afford a closed form analytical solution. Blumberg therefore catalogued
analytic expressions (when an explicit integration can be carried out) of the growth function N(t)
for various values of the parameters and .
The population at the inflection point, Ninf , is given by

N inf =

K
+

This also coincides with that of the Verhulst logistic equation when = . For >> the
inflection occurs very near the carrying capacity, and for << , Ninf approaches 0 and inflection
occurs only if N0<Ninf .
Figure 5 and figure 6 exhibit respectively the population size as a function of time and the
growth rate variation with population size for several values of the parameters and . In
Figure 6 the gradual transition of the inflection point from values less than

< , to values greater than

K
, when > , can be clearly seen.
2

K
= 50 , when
2

30

R.L.I.M.S. Vol. 2, May 2001

5. Blumberg's logistic curve

100

90
alpha=1.5
gamma=2.5

80

alpha=1.5
gamma=1.5

Population size N

70

60

50

40

alpha=2.0
gamma=1.5

30

10

alpha=2.0
gamma=1.0

alpha=1.0
gamma=3.0

20

20

40

60

80

100

120

Time t

Figure 5. Population size growth versus time according to Blumbergs functional form.
6. Blumberg's equation
18
alpha=1.0
gamma=3.0

16

alpha=2.0
gamma=3.0

14

alpha=1.5
gamma=2.5

Growth rate dN/dt

12

10
alpha=3.0
gamma=3.0
8

6
alpha=4.0
gamma=2.0
4

0
10

20

30

40

50

60

70

80

90

100

Population size N

Figure 6. Growth rate versus size graph for Blumbergs equation. The movement of
inflection point from the left, where < , to the right, where > , is clearly visible

the

3.3. Von Bertalanffys growth equation

Von Bertalanffy [13] introduced his growth equation to model fish weight growth. Here the
Verhulst logistic growth curve was modified to accommodate crude metabolic types based upon
physiological reasoning. He proposed the form given below which can be seen to be a special
case of the Bernoulli differential equation:

A. Tsoularis, Analysis of Logistic Growth Models

31

2
3
dN
N

3
= rN 1
K
dt

which has solution

1

1 rK 13 t
3
N

N (t ) = K 1 + 1 0 e 3

The Bertalanffy model cannot be derived from the Turner model as the values of the exponents,

2
1
= , = , = 1,
3
3

violate the condition

= 1 + (1 )

stipulated by Turner et al. (see

Section 3.5). It cannot therefore be seen as a special case and should be viewed as a separate model
accordingly.
Here, Ninf is given by
3

N inf

K= 8 K
=
27
2 1
+
3 3

which, whilst differing from that for the Verhulst curve, still represents a substantial restriction for general
modelling purposes.
Figures 7 and 8 display respectively a typical Von Bertalanffy weight growth curve and its
inflection.
7. Von Bertalanffy's growth curve
100

90

80

Weight N

70

60

50

40

30

20

10

20

40

60
Time t

Figure 7. Von Bertalanffys weight growth curve.

80

100

120

32

R.L.I.M.S. Vol. 2, May 2001

8. Von Bertalanffy's curve

1.6

1.4

Weight growth rate dN/dt

1.2

0.8

0.6

0.4

0.2

0
10

20

30

40

50

60

70

80

90

100

Weight N

Figure 8. The growth rate for Von Bertalanffys form realizes its maximum at around 30, when
the asymptotic value is K = 100.
3.4. Richards growth equation
Richards extended the growth equation developed by Von Bertalanffy to fit empirical plant data
[12]. Richardss suggestion was to use the following equation which is also a special case of the
Bernoulli differential equation

N
dN
= rN 1
dt
K

(6)

which has the solution

N 0
rt
N (t ) = K 1 e 1

Unlike its Von Bertalanffy antecedent however, the Richards growth function does follow from the
Turner model (Section 3.5) in the case where, = 1.
Here inflection occurs at
1

N inf

1
K
=
1
+

For = 1, (6) trivially reduces to the Verhulst logistic growth equation (1). For extreme values
of we obtain the following values for Ninf :

lim N inf = Ke 1
0

lim N inf = K

Consistent with the previous observation, the above values also follow from the corresponding
population value at the inflection point for the generic growth function with = 1.

A. Tsoularis, Analysis of Logistic Growth Models

33

Figure 9 illustrates four different Richards growth curves with = 0.01, 0.05, 3.0 and 6.0. Figure
10 displays the variation of the weight growth rate for small values for , = 0.01 and = 0.05
giving inflection at Ninf 36 and Ninf 38 (both values are approximately equal to
larger values resulting in higher inflection values.
9. Richards' Growth Equation
100

beta=0.05

90

80
beta=6.0

Plant weight N

70

60

50
beta=3.0
40

30

20

10

beta=0.01

20

40

60

80

100

Time t

Figure 9. Plant weight growth in time according to Richards equation.

120

K
), and
e

34

R.L.I.M.S. Vol. 2, May 2001

10. Richards' Growth

140
beta=3.0
120

100

Growth rate dN/dt

beta=6.0

80

60

beta=0.05

40

20
beta=0.01
0

10

20

30

40

50

60

70

80

90

100

Plant weight N

Figure 10. Weight growth rate versus weight plot for Richards equation.
3.5. Gompertz growth function
The Gompertz growth curve can be derived from the following form of the logistic equation as
a limiting case:

N
dN
r
= N 1
dt

rN
=
K

K N

K N
= r N

r
.
K
K N
in the limit as 0 we obtain
Looking at

where

r =

A. Tsoularis, Analysis of Logistic Growth Models

35

K N
e ln K e ln N
= lim
0
0

lim

( ln K ) n ( ln N ) n

n!
n!
n =0
n=0
= lim
0

n 1

K
= ln + lim
(ln K ) n (ln N ) n

N
n = 2 n!
K
= ln
N
Similarly, lim (r ) = r , > 0.

The growth rate modelled by the Gompertz function is given by

K
dN
= rN ln
dt
N

(7)

With > 0, 1 , this special case is more usually known as the hyper-Gompertz (Turner et al.
[17]), generalized ecological growth function, or simply generalized Gompertz function.
Equation (7) can be conveniently rewritten as follows

N
d N
ln = r (1) ln

dt K
K

which upon integration leads to the analytic solution

N 1
1

0
N (t ) = K expln
+ r (1) (1 )t
K

inflection point, Ninf , is obtained by differentiating both sides of

The population at the

K N
dN
= r N

dt

d 2N
, setting
= 0 , and subsequently taking the limit:
dt 2

1
N inf = lim
0 1 +

K = Ke

The above form can also be obtained by differentiating both sides of equation (7) and setting

d 2N
= 0.
dt 2

For relatively large increasing positive values of the inflection point tends to 0. Indeed, to
achieve an inflection point, it is necessary that

Ke > N 0 , whereas for increasingly small values

of , the inflection point tends to K.

For = 1 the equation

K
dN
= rN ln
dt
N

is the ordinary Gompertz growth (see [19], [20]). The solution to (8) is

(8)

36

R.L.I.M.S. Vol. 2, May 2001

N
N (t ) = K expln 0 e rt

K
1
The population value at the inflection point, N inf = Ke , is obtained from that for the
generalized Gompertz growth function with =1.
Figure 11 depicts three hyper-Gompertz growth curves ( > 0, 1) and the ordinary Gompertz
growth ( = 1) for N0 = 10 and K = 100. As predicated, relatively large values of result in an
inflection near the origin, = 3 gives Ninf 4.5 < N0, = 2.5 gives Ninf 7.5 < N0, whereas
relatively low values of result in a visible inflection point. A value of = 0.5 gives Ninf 60, and
= 1 gives Ninf 35.5, as shown in Figure 12.
11. Hyper-Gompertz Growth Function
100
gamma=0.9
gamma=1.0

90

gamma=2.5
80

Weight N

70

gamma=5.0

60

50

40

30

20

10

Time t

Figure 11. Hyper-Gompertz and ordinary Gompertz ( = 1) growth plots.

A. Tsoularis, Analysis of Logistic Growth Models

37

12. Hyper-Gompertz Growth

250

Weight growth rate dN/dt

200

gamma=3.0

150

100

gamma=2.5
gamma=0.5

gamma=1.0
50

0
10

20

30

40

50

60

70

80

90

100

Weight N

Figure 12. Hyper-Gompertz and ordinary Gompertz ( = 1) weight growth

curves.

rates versus weight

4. Generalized Logistic growth function

Here we propose a generalized logistic growth equation which incorporates all previously reported
functional forms as special cases. We will adopt the term generalized logistic equation in our
exposition, a term first used by Nelder [14] to describe the Richards equation. We believe the
adopted term is an appropriate one as it connotes exactly what it purports to achieve.
4.1. Definition and properties of the generalized logistic function
We define the generalized logistic function thus

N
dN

= rN 1
dt
K

(9)

where , , are positive real numbers. In this paper we confine ourselves to positive values
for these parameters and that for r, as negative exponents do not always provide a biologically
plausible model. Unlike Lotkas derivation of the Verhulst logistic growth equation from the
truncation of the Taylor series expansion of f(N) near N = 0, (9) cannot be derived from such
an expansion unless , , are all positive integers, in which case a power series can be
( 1)

(0) = 0 .
generated with the first terms f (0) = f (0) = f (0) =  = f
By differentiating (9) and setting the second derivative to zero, we obtain the
parametric expression for the population value, Ninf , at the inflection point
N inf

= 1 +

following

Clearly if Ninf < N0 , no inflection is possible as the population will have started with this initial value,
N0, and with a positive intrinsic growth per capita rate thus ensuring that Ninf is not achievable.

38

R.L.I.M.S. Vol. 2, May 2001

The above expression can be seen to contain the inflection points of all previous curves as
special cases. For example, with = 1 + (1 ) , it reduces to the inflection value for the
generic growth model given by (4).

dN
= rN
dt

Exponential Growth

=1, = 0

Logistic Growth

Von Bertalanffy
1

2
dN
N 3
3
= rN 1
K
dt

2
1
= , = , = 1
3
3

dN
N

= rN 1
dt
K

== =1

N
dN

= rN 1
dt
K

Generalized Logistic

= 1 + (1 )
N
dN
= rN 1+ (1 ) 1
dt
K

=1Blumbergs Equation

N
dN
= rN 1
dt
K

Generic Growth

N
dN
= rN 1
dt
K

K N
dN
= lim r N

dt 0

==1

Richards Equation

=1

K
dN
= rN ln
dt
N

K
= rN ln
N

Hyper-Gompertz Function

Gompertz Function
Figure 13. The generalized logistic curve and its derivative models.
Figure 14 is a display of the generalized logistic growth curve with parameters , , chosen at
random (but kept positive), and figure 15 shows the growth rates with their respective maxima
for the same parameter range.

A. Tsoularis, Analysis of Logistic Growth Models

39

14. Generalized Logistic Growth Curve

100

90

80

alpha=0.5
beta=0.9
gamma=1.5

Population size N

70
alpha=1.0
beta=0.5
gamma=2.0

60

alpha=1.0
beta=2.0
gamma=3.0

50

40

alpha=0.5
beta=1.0
gamma=1.5

30
alpha=2.0
beta=3.0
gamma=4.0

20

10

20

40

60

80

100

120

Time t

Figure 14. The generalized logistic growth rate curve for several parameter triplets (,,) and X0
= 10, K = 100.

15. Generalized Logistic Growth

7

alpha=2.0
beta=3.0
gamma=4.0

alpha=1.0
beta=0.5
gamma=2.0

Growth rate dN/dt

5
alpha=1.0
beta=2.0
gamma=3.0
4

alpha=0.5
beta=0.9
gamma=1.5

0
10

20

30

40

50

60

70

80

90

100

Population size N

Figure 15. Population growth rate as function of population size according to the generalized
logistic form.

40

R.L.I.M.S. Vol. 2, May 2001

N
By introducing the auxiliary variable x = we can transform the autonomous differential
K

equation (13) into

dx
= rK 1 x
dt

1
+1

(1 x)

and upon separation of variables and subsequent integration from 0 to t

N (t )

N0

1
1

(1 x) dx = rK 1t

(10)

For a certain configuration of parameters the above integral is the Incomplete Beta Function [21].
Specifically, the parameter configuration (assuming again all parameters are positive) is

0 < < 1, > 0, 0 < < 1

(1 x) term in the integrand of (10) step by step integration

1
can be performed. For notational convenience let =
1 . Then

By expanding binomially the

N (t )
x =

x
( + i 2)! x
1
( 1)! ( 1)!( + ) N = rK t

i
i
i =1

x = 0
i

(11)

The above form provides an analytical expression of t as a function of

x=

N
. An alternative,
K

but equivalent, infinite series expression to (11), taken from Abramowitz and Stegun [22], is
reported in the book by Banks [23]. The more desirable functional form,

x(t ) =

N (t )
, is one
K

of the real roots of the polynomial arising from truncating (11) thus,
n

a x
i

+i

+ g (t ) + O( x + n +1 ) = 0

i =1

where

g (t ) = rK

t + constant and ai = ai(, ), are coefficients functionally dependent on

and .
4.2. Comments on the inflection value Ninf
In order to assess properly the evolution of the inflection value,

N inf

= 1 +

K , with

the simultaneous variation of the parameters , , we have created surface plots of Ninf. On
each surface plot one parameter is held constant and the other two are allowed to vary.
Figure 16 is a three-dimensional plot of Ninf versus parameters , . Here as well as and

are allowed to vary continuously in the range [1.0 , 100.0]. It can be seen that N inf as

This is because

N inf

K
= 2 1 +

1
1

>0

A. Tsoularis, Analysis of Logistic Growth Models

41

= = = 100.0 , N inf K = 100 . For = = 100.0 and = 10.0 , Ninf 50,

whereas for = = 100.0 and = 10.0 , Ninf 100. For much smaller values of , say =
10.0 we have a similar situation: for = 100.0, = 10.0 , Ninf 100, and for
= 10.0, = 100.0 , Ninf 10. This indicates that for relatively small values of , large
For

values of have a moderating effect on the growth of the inflection value regardless of the
value of .

Figure 16. Surface plots of the inflection value, Ninf (K = 100) as a function of two variables ,.
Each surface corresponds to a particular value of the parameter
Figure 17 is a three-dimensional plot of Ninf versus parameters

. Here again , , are

allowed to vary continuously in the range [1.0 , 100.0]. Again it can be seen that

This is because

N inf

K
1 +

ln1 +
1 +

>0
The resulting inequality is established by observing that the following inequality

ln y >
where

y = 1+

y 1
for y > 1,
y

, is always true given the restrictions on , and .

N inf as

42

R.L.I.M.S. Vol. 2, May 2001

It is obvious from the plot that is the dominant parameter in that large values of result in a
high inflection value, Ninf > 90, regardless of the magnitude of the other two parameter values.
For small values of however, say = 0.5, is the dominant parameter in that small values of
will drive the inflection value to zero and large values of will result in a rapid increase of
the inflection value.

Figure 17. Surface plots of the inflection value, Ninf (K = 100) as a function of two variables ,.
Each surface corresponds to a particular value of the parameter in the range [0.50 , 100.0].
Figure 18 is a three-dimensional plot of Ninf versus parameters

. Here also , , are

allowed to vary continuously in the range [1.0 , 100.0]. In this case however,

N inf as

because

N inf

K
= 1 +

<0

1
1

Very small values of and rapidly drive the inflection value to zero regardless of the value
of . Large values of either or , or both, will restore the inflection value to its asymptotic
value when is relatively small, say = 10.0. When is large however, say = 100.0, large
values of and small values of result in a much more rapid growth of the inflection value
towards K than correspondingly large values of and small values of , which produce a
maximum inflection value of 50.

A. Tsoularis, Analysis of Logistic Growth Models

43

Figure 18. Surface plots of the inflection value, Ninf (K = 100) as a function of two variables
,. Each surface corresponds to a particular value of the parameter in the range [10.0 ,
100.0].
It is clearly evident from inspection of figures 16, 17 and 18 that for intermediate to large values
of , , (typically >10, > 40, > 40 for figures 16 and 17 and >40, > 40, > 10 for
figure 18) the corresponding surfaces consistently exhibit approximately zero curvature, that is,
they are effectively planes. This is seen immediately from considering the total differential dNinf
along any surface, for example, = constant (exhibited in figure 16):

dN inf =

N inf

d +

N inf

We are going to prove that dNinf vanishes identically on the surface = constant by proving
that

N inf

>> 1

and

N inf

vanish identically for relatively large values of

we shall use the approximation

N inf

1+

K
= 1 +

1
1

1
1

1
K

=
+1

and

Since

44

R.L.I.M.S. Vol. 2, May 2001

Then

lim

Next consider the limit of

1
1

= lim K
,

N inf

N inf

1
+1

=0

after rearranging terms:

N inf

K

= 1 +

ln1 +

1 +

1 1
ln =

ln

1
1+

Then

lim

N inf

K
,

lim

1+

ln
lim
,

since

lim

ln y

= 0 , where y =
, and
y

lim

=0.

This directly implies that for an infinite configuration of any two parameters and a fixed (but
relatively high) value of the remaining parameter, the time needed by the population (or growing
organism) to reach a given inflection value, tinf , is not unique, or equivalently, that a given
inflection value, Ninf , can be attained via many different parameter values and hence at different
times tinf .

5. Discussion
The classical Logistic growth equation of Verhulst has been used a basis for several extended models.
Each is shown to accommodate population or physical growth without the restriction that the maximum
growth rate must, rather artificially, occur at half of the carrying capacity in the case of population
dynamics, or maximum attainable dimension for physical growth, for the system being modelled.
Properties and restrictions previously unreported for these models are identified and presented.

A. Tsoularis, Analysis of Logistic Growth Models

45

A generalized form of logistic growth (13) has been introduced which encompasses extended logistic
growth models as special cases. We have shown that the general solution to (13) is an
Incomplete Beta Function which has been tabulated by Pearson [21]. In addition we have dealt
rigorously, for the first time, with the population at the inflection point, Ninf , of all the growth
models presented. Furthermore, for this generalised form, we have proved that for extreme values of
certain parameters, the population will reach a fraction of the carrying capacity, K. This is a
novel feature possessed by the generalized logistic growth curve alone and differentiates it from the
Verhulst logistic growth curve (1) and the extended forms which have been discussed. It also
emphasizes its modelling utility in situations where a biological population (or growing organism) is
unable, for whatever reason, to reach the expected carrying capacity of the environment (or
limiting dimension).

References
[1] P.F. Verhulst, Notice sur la loi que la population suit dans son accroissement, Corr.
Math. Physics, 10 (1838) 113.
[2] T.Carlson, ber geschwindigkeit und grsse der hefevermehrung in wrze, Biochem. Z,
57, (1913), 313-334.
[3] R.Pearl, The growth of populations, Quarterly Review of Biology, 2, (1927), 532-548.
[4] R.Pearl, Introduction of Medical Biometry and Statistics, Saunders, Philadelphia, 1930.
[5] B.J.T. Morgan, Stochastic models of groupings changes, Advances in Applied Probability,
8, (1976), 30-57.
[6] C.J.Krebs, The Experimental Analysis of Distribution and Abundance, Harper and Row,
New York, 1985.
[7] T.C.Fisher, R.H.Fry, Tech. Forecasting Soc. Changes, 3, (1971), 75.
[8] C.Marchetti, N.Nakicenovic, The Dynamics of Energy Systems and the Logistic
Substitution Model, Int. Inst. for Appl. Sys. Anal., Laxenburg, Austria, 1980.
[9] R.Herman, E.W.Montroll, Proceedings of the National Academy of Sciences, USA, 69,
(1972), 3019.
[10] A.J.Lotka, Elements of Mathematical Biology, DOVER, New York, 1956.
[11] R.Pearl, L.J.Reed, On the rate of growth of the population of United States since 1790
and its mathematical representation, Proceedings National Academy of Sciences USA,
G (1920), 275.
[12] F.J.Richards, A Flexible Growth Function for Empirical Use, Journal of Experimental
Botany, 10(29), (1959), 290-300.
[13] L.Von.Bertalanffy, A quantitative theory of organic growth, Human
Biology, 10(2),
(1938), 181-213.
[14] J.A.Nelder, The fitting of a generalization of the logistic curve, Biometrics, 17, (1961),
89-110.
[15] A.A.Blumberg, Logistic Growth Rate Functions, Journal of Theoretical Biology, 21,
(1968), 42-44.
[16] M.E.Turner, B.A.Blumenstein , J.L.Sebaugh , A generalization of the logistic law of
growth, Biometrics, 25, (1969), 577-580.
[17] M.E.Turner, E.Bradley, K.Kirk, K.Pruitt, A Theory of Growth, Mathematical Biosciences,
29, (1976), 367-373.
[18] R.Buis, On the Generalization of the Logistic Law of Growth, Acta Biotheoretica, 39,
(1991), 185-195.
[19] J.W.Haefner , Modeling Biological Systems, ITP, 1996.
[20] H.McCallum, Population Parameters: Estimation for Ecological Models, Blackwell
Science, UK 2000.
[21] K.Pearson, Tables of the Incomplete Beta Function, Library Binding, Lubrecht and
Cramer Ltd., 1968.

46

R.L.I.M.S. Vol. 2, May 2001

[22] M.Abramowitz, I.M.Stegun, Handbook of Mathematical Functions, Dover, New York,

1965.
[23] R.B.Banks, Growth and Diffusion Phenomena, Springer Verlag, 1994.