What is mathematical modelling and why is it so important

Introduction to Mathematical modelling

Before attempting to define mathematical modelling, let us first try
A.G. Fareo
answer the question on our minds right now, which is
School of Computer Science and Applied Mathematics What exactly is the point of mathematical modelling?
University of the Witwatersrand
Johannesburg A simple answer to that is, ........................ to understand our
world, and hopefully, to make it better.
Modelling III
Our world is filled with physical processes, some man-made, and
some are not man-made. .

2
Physical processes (MAN-MADE OR NATURAL) Physical processes (MAN-MADE OR NATURAL)

Examples of man-made processes are:

Industrial processes, e.g., mining, fracking, petroleum/gas Example of natural processes are:
Climatic processes such as hurricanes, flooding, tsunamis; seismic
processing; factory processes e.g., sugar extraction from sugarcane,
processes such as earthquakes; physiological processes such as
glass moulding/ blow moulding/extrusion, batch reactors; blood flow in arteries, tumour growths, cell growth, cell
environmental processes, e.g., pollution, carbon emission, etc; proliferation and mutation, etc; enviromental processes e.g.,
spread of diseases (hiv, corona virus, hpv, sti’s); financial processes ecology, wildlife conservation
.
such as amortisation, annuity, etc.
.

3 4
What is Mathematical modelling Industrial Mathematics ( or Mathematics in industry)

Mathematical modelling is simply the use of mathematics to

A branch of applied mathematics which focuses on problems from
understand and explain a real world problem.
industry. These problems are presented to applied mathematicians
It involves
by industry representatives who know the problem and want
→ Dissecting/deconstructing/analysing the real world problem
solutions to the problem. Most times these industry representatives
→ using mathematics to phrase/ narrate the real world problem are engineeers, or scientists. The aim of industrial mathematics is
→ simplify the mathematical model using commonsense to study the industrial problem, solely using the tools of
assumptions mathematics, with the goal of pro!erring solutions which are
→ solving the model to obtain results that can be tested against relevant to the industry.
what is observed in real life Industrial mathematician can find employment in the following
industry: signal processing, computer graphics, risk management,
Most real world problems that are modelled using mathematics
production line optimization, software testing and reliability, etc. .
come from the industry. .

5 6
Basic steps in modelling Population models

We will attempt to model population dynamics using continuous

models. When thinking about it, it may sound impossible to model
→ Problem Identification
the growth of any species using di!erential equations. This is
→ Model formulation
because the population P(t) of any species changes all the time by
→ Model reduction
→ Model analysis integer amounts, and therefore cannot be a di!erential function of
→ Computation time.
→ Model validation If, however, P(t) is large and say, it increases or decreases by a
very small amount, then, the change is small compared to a given
population.
.

7 8
Population models Population models
It is reasonable to assume the following:
The assumption of continuity and di!erentiability of P(t) with
respect to time provides a reasonable approximation. The use of
number of births in a short time interval ↓ population at the
time as a real continuous variable is justified using similar
beginning of this interval and to the length of the interval.
arguments.
That is,
Let P(t) denote the size of a population of a given isolated species
number of births = ω(t , P(t))P(t)ωt
at time t and let ωt be a small time interval. The population at
any time t + ωt will satisty

number of deaths in a short time interval ↓ population at the

P(t +ωt)↑P(t) = number of births in time ωt ↑number of deaths in time ωt
beginning of this interval and to the length of the interval.
That is,
.
number of deaths = µ(t , P(t))P(t)ωt
9 10
.
Population models Exponential growth population models
Thus,
P(t + ωt) ↑ P(t) = (ω ↑ µ)P(t)ωt The simplest form of r (t , P(t)) occurs when

Let r (t , P(t)) = (ω ↑ µ). Therefore, r (t , P(t)) = r , where r ↗ R.

P(t + ωt) ↑ P(t) = r (t , P(t))P(t)ωt The resulting population model given by

Dividing through by ωt, and taking the limit as ωt ↔ 0 yields dP

= rP(t). (1)
dt
dP
= r (t , P(t))P(t). is used in a short term population forecasting.
dt
.
In order to proceed, the form of r (t , P(t)) has to be specified.
.
11 12
Exponential growth population models Exponential growth population models
The solution to
dP
When we divide the rate of change of population by the population = rP(t).
dt
in (1), we get the relative growth rate
is
dP(t) P(t) = P(t0 )e r (t ↑t0 ) .
r = dt .
P(t) where P(t0 ) is the initial size of the population, which is at time t0 .
If population is measured in people, and time in years, then Example 1:
Assume that a population of P = 200 million grows at a relative
people/year 1 % of population
r= = = growth rate of 3% per year. This means that the population
people year year
increases at the rate
.
dP
= rP(t) = 0.03 ↘ 200million = 6million/year.
dt
13 14
Exponential growth population models . test the accuracy of this result, let us calculate when the Earth
To
population is expected to double.
Example 1: Let t = 0 in the year 1965. Thus
According to the data from the US department of commerce, It
was estimated that the earth population in 1965 was 3.34 billion 2P(0) = P(0)e 0.02t
and that the population was increasing at an average rate of 2%
Solving yields
per year during the decade 1960 - 1970.
t = 50ln2 ≃ 34.6years
Thus, P(t0 ) = 3.34 ↘ 109 and r = 0.02. The solution P(t) takes the
form In about 35 years later, the model predicts population will be 6.68

P(t) = 3.34 ↘ 109 e 0.02(t ↑1965) billion. The estimated population in the year 2000 was 6.114
billion. There is of course a good agreement, despite that
.
overestimate.
.
15 16
comparison: actual population and model population

The exponential model ignores the fact that there will be

overcrowding as population increases, and therefore, the individual
members will have to compete with each other for the limited
available living space. Also, no matter how plentiful food is, sooner
or later, the population will start running out of food and other
resources as the population grows exponentially. When resources
become limited, birth rate and death rate is a!ected. Birth rate
will actually decrease. Death rate may actually increase.
.

.
17 18
Logistic equation for Population model
The simplest way to take the aforementioned into account is to
P
assume r (t , P(t)) = r (1 ↑ K ).
It is reasonable to think that a given habitat can sustain only a
The resulting population model given by
finite number K of individuals called the carrying capacity. And the
! "
closer the population is to this number, the slower is it growth. As dP P
= rP 1 ↑ . (2)
the population reaches its limiting value, the rate of the population dt K

growth, although still increasing, approaches its limiting value of which proved to be one of the most successful models for
zero. describing a single species population.
. Here the relative growth rate dP
dt /P is no longer constant; it now
depends on the population so that, as the population P increases,
the relative growth rate decreases. We will use this di!erential
equation in a di!erent context later on, in epidemics.

19 . 20
Logistic equation for Population model comparison: real solution vs model solution
Take experimental value K = 10.76 billion, r = 0.029, P0 = 3.34
The solution to the ode
billion at t = 1965
! "
dP P 3.34 ↘ 109 (10.76 ↘ 109 )
= rP 1 ↑ . (3) P(t) =
dt K
3.34 ↘ 109 + (10.76 ↘ 109 ↑ 3.34 ↘ 109 )e ↑0.029(t ↑1965)
subject to P(t0 ) = P0 is

P0 K
P(t) = (4)
P0 + (K ↑ P0 )e ↑r (t ↑t0 )
Notice that
lim P(t) = K .
t ↔⇐

Thus, our model accurately points to our earlier discussion that a

given habitat can only sustain up to a finite number K of
individuals.
.
21 22
.
Logistic equation for Population model Logistic equation for Population model

Substitute (4) into (3) in order to show a relationship between

K , P0 , r and dP/dt:
dP rP0 K (K ↑ P0 )e ↑r (t ↑t0 )
=# $2 (6)
dP rP0 K (K ↑ P0 )e ↑r (t ↑t0 ) dt P0 + (K ↑ P0 )e ↑r (t ↑t0 )
=# $2 (5)
dt P0 + (K ↑ P0 )e ↑r (t ↑t0 ) If P0 > K , dP/dt < 0 for a finite time t and the population
decreases monotonically. However, this decrease is shortlived since
If P0 < K , then dP/dt > 0 for a finite time t and the population
as t gets bigger and bigger, dP/dt ↔ 0. The decreasing population
increases monotonically. However, this increase is shortlived since
P(t) then slowly converges to the carrying capacity K .
as t gets bigger and bigger, dP/dt ↔ 0. Population P(t) then
.
converges to the carrying capacity K .
.

23 24
 Take experimental value K = 10.76 billion, r = 0.029, t0 = 0 Example: Spread of a disease in a community

You are asked to model the spread of a disease in a community.

This task is quite a vague one.
What are the important questions to ask and assumptions to
make?

.
25 26
Example: Spread of a disease in a community Example: Spread of a disease in a community

You are asked to model the spread of a disease in a community. You are asked to model the spread of a disease in a community.
This task is quite a vague one. This task is quite a vague one.
What are the important questions to ask and assumptions to What are the important questions to ask and assumptions to
make? make?
→ Is the community isolated? → Is the community isolated?

→ How many people are in the community?

27 28
Example: Spread of a disease in a community Example: Spread of a disease in a community

You are asked to model the spread of a disease in a community. You are asked to model the spread of a disease in a community.
This task is quite a vague one. This task is quite a vague one.
What are the important questions to ask and assumptions to What are the important questions to ask and assumptions to
make? make?
→ Is the community isolated? → Is the community isolated?

→ How many people are in the community? → How many people are in the community?

→ Are people currently sick? If so, how many? → Are people currently sick? If so, how many?

→ Is the disease bad and deadly?

29 30
Example: Spread of a disease in a community Example: Spread of a disease in a community

You are asked to model the spread of a disease in a community. You are asked to model the spread of a disease in a community.
This task is quite a vague one. This task is quite a vague one.
What are the important questions to ask and assumptions to What are the important questions to ask and assumptions to
make? make?
→ Is the community isolated? → Is the community isolated?

→ How many people are in the community? → How many people are in the community?

→ Are people currently sick? If so, how many? → Are people currently sick? If so, how many?

→ Is the disease bad and deadly? → Is the disease bad and deadly?

→ Is there a vaccination? → Is there a vaccination?

→ How do the people get sick?

.
31 32
Infection models Example: Spread of a disease in a community

Consider two distinct populations, the healthy (susceptible) and

the sick (infective), and call this the SI epidemic model. We will
Infection spreads when healthy people "catch" the disease from
model the spread of the disease using Compartmental modelling.
sick people. When a healthy person comes into contact with a sick In this simple model, no birth and no death. No immigration.

person, sometimes the healthy person gets sick. Notice that when

two healthy people or two sick people come in contact, nothing

happens: healthy people cannot infect others and the sick people

cannot become sick again. Let S(t) be the number of healthy people at time t. Let I(t) be the
. number of sick people at time t. Denoting the total population by
N, we see that N = S + I.
.

33 34

Each person can only move from the susceptibles to the infectives
compartment. Thus, as the number of susceptibles decreases, the Number of contacts you chill with (contact rate c)

number of infectives will increase with time.

Let c = c(N) be the number of contacts made by a susceptible
Thus
person per unit time (e.g per day) (not su"cient for transmission )
Rate of change in the susceptible population = zero - rate at
What do we mean by contacts?
which people get sick and as a result leaving the compartment.
Viral diseases (e.g corona viruses, influenza, etc ): Contacts are the
What is the rate at which people get sick (infection rate)?, given people you meet with who are within close proximity to you.
that total population is N, and that Viral diseases (e.g. HIV) : Contacts are the people you have as
sexual partners.
I(0) = I0 , S(0) = N ↑ I0 Viral diseases (Ebola) : Contacts are the people whose bodily
fluids (blood, faeces, vomit) one has had contact with.
.
.
35 36
 Risk of infection (Transmission risk p)
Therefore, the number of contacts which is su"cient for
This has to do with contact between an infectious individual and a
susceptible individual and the chance that the susceptible gets transmission per susceptible per time is pc = ω(N).

infected by the infectious individual as a result of the contact.

The number of contacts su"cient for transmission per time
Let the probability of a susceptible getting infected for every
=e!ective contact rate = disease transmission rate
contact made with an infectious individual be p.

Thus, e!ective contact rate per susceptible (disease transmission

For HIV,p = 0.04% (for insertive penile-vagina intercourse ) and
p = 0.08% (for receptive penile-vagina intercourse ) according to rate per susceptible) = cp = ω(N)

CDC (US). .
What is p for influenza and corona virus?
.
37 38

Proportion of these contacts that are infective

Standard Incidence: contact rate is a constant ⇒ ω(N) = ω, ω ↗ R per susceptible person per time is equal to prevalence of infectives
↘ number of contacts su"cient for transmission = NI ↘ ω(N).
Mass action Incidence: contact rate increases linearly with

population size ⇒ ω(N) = ωN , ω↗R Thus, ω(N) IS

N is the total number of contacts that are infective per
unit time. It is therefore the rate of infection, or the number of
.
new cases per unit time.
.

39 40
The rate of infection is the rate at which healthy people get sick, Governing Equation:
for example, per day.
The contact rate, ω, may be chosen as a constant, for which,
dS S(t)I(t)
regardless of the population size, the number of people we are in = ↑ω(N)
dt N
contact with remains the same. dI S(t)I(t)
= ω(N)
dt N
IS
number of new cases per time = ω(N)
N Let ω(N) = ωN(mass action incidence), ω ↗ R, then

dS
= ↑ωSI
dt

dI
= ωSI
dt

to be solved subject to I(0) = I0 and S(0) = N ↑ I0 .

.
41 . 42
The solution to the nonlinear system is given by

I0 N
I(t) = (7)
(N ↑ I0 )e ↑N ωt + I0

and
S(t) = N ↑ I(t)

The long time behaviour of I(t) and S(t) can be obtained:

As t ↔ ⇐
I(t) ↔ N

and
S(t) ↔ 0.
Figure: N=30 and ω = 0.035
.
43 44
 In the SI model, once a susceptible person becomes infected, he or
From the graphs , we see that the infectious population curve she will remain infective forever. This applies to epidemics with no
asymptotically approaches the total population, N. That is, immunity, e.g., AIDS.
I(t) ↔ N as t ↔ ⇐, and that, S(t) ↔ 0 and t ↔ ⇐.
Also, as I0 increases, it is taking less longer for S(t) to decay to
zero. More and more people are also getting infected per time. For ω I(1)
0.5 30
example, at t = 4, the table below shows the number of people who
0.2 28
are infected as I0 increases.
0.1 12
0.05 4
I0 1 2 5 10
I(4) 21 25 28 30
. Figure: I0 = 1 and N = 30
.

45 46
SIR model
A more realistic model is the SIR model also called the
Kermack-Mckendrick SIR model. This model accounts for three
distinct populations: the healthy (Susceptibles), the sick Mass action incidence (ω = ωN)
(Infectives), and the recovered (Removeds). In 1927, Kermack and
dS
= ↑ωS(t)I(t),
McKendrick proposed a system of di!erential equations for the SIR dt
model. dI
= ωS(t)I(t) ↑ µI(t)
Model subdivides total population at time t, denoted N(t) into dt

susceptible S(t), infected I(t) and recovered R(t) compartments. dR

= µI(t)
dt
Thus
subject to S(0) = 29, I(0) = 1 and R(0) = 0.
N(t) = S(t) + I(t) + R(t)
ω is the average number of contacts an infective makes per unit
time. 1/µ is the mean duration of infectivity.
.
.
47 48
In the SIR model, recovereds can no longer infect susceptibles nor
can they become infected themselves. The initial conditions are
the same as in the SI model, with zero recovereds at the
beginning. The total population is the sum of the three distinct
populations: N = S + I + R.

Please note: The removal rate is taken to be proportional to the

infective population. Figure: SIR model with ω = 0.035 and µ = 0.1

.
The graph of S(t), I(t) and R(t) are shown. Initially, there are 29
susceptibles, one infective and zero recovereds. And as time
evolves, the number of susceptibles goes to zero while the number
of infectives rises up to a peak and then begins to drop.
49 . 50
Model with birth and death
Let P(t) represent the number of healthy people at time t.
Let I(t) be the number of sick people at time t.
Let
N = P(0) + I(0)

We will model the spread of the disease using Compartmental

(a) (b) modelling.

Figure (a): As disease transmission rate ω decreases, the rate of

growth of the infected population decreases, flattenning out the
curve.
Figure (b): As the rate of recovery from disease increases, fewer
and fewer people remain infectious.

51 52
 Governing Equations

Initially,
N 4N
I(0) = , P(0) =
5 5

The disease spread through contact and interaction between

healthy and sick people. The healthy people are susceptible.
.
! "
dP P
= εP 1 ↑ ↑ ωPI
dt K
dI
= ωPI ↑ ϑ I
dt

.
53 54

55 56
Exercise 1 Exercise 1 continued

(1.) Solve the equations for I(t) and P(t) subject to (4.) Did you notice any conservation principle used in the
I(0) = N/5, P(0) = 4N/5. Provide plots in Matlab or modelling?
Mathematica. Experiment with di!erent parameter values for (4.) Ideally, the population of any species always change by integer
ε, ω, ϑ and N. amounts with time. Also, time should be discrete, not
(2.) Suppose a vaccination is now available. People in P get continuous. Therefore population and time need not to be a
vaccinated at a rate ϖP. Rewrite the governing equations for continuous variable. Population should therefore never be a
this scenario and plot your results in Mathematica. di!erentiable function of time. On what grounds am I able to

(3.) Discuss on the two model above consider the population problem as continuous?

. .

57 58

Even though population density and contact rate appears to be

increasing with population size, it may be fair to think that for
humans, contact rate is weakly dependent on population size. It
SIR model (No vital dynamics)
can be assumed that the daily contact patterns of people do not
change, in large or small communities. For example, students of dS S(t)I(t)
= ↑ω(N) ,
the same age group in a country play together and as such, they dt N
have a similar number of daily contacts. dI S(t)I(t)
= ω(N) ↑ µI(t)
. dt N
dR
= µI(t)
dt

59 60
 Standard Incidence: ω(N) = ω
Divide S , I and R by the population size N. That is,
S = S/N , I = I/N , R = R/N. Resulting model (after dropping
the overbar) dS
= ↑ωS(t)I(t),
SIR model (No vital dynamics) dt
dI
= ωS(t)I(t) ↑ µI(t) (8)
dS dt
= ↑ωS(t)I(t),
dt
The region of interest is given by
dI
= ωS(t)I(t) ↑ µI(t) %
dt %
T = {(S , I)%%S ⇑ 0, I ⇑ 0, S + I ⇓ 1}
dR
= µI(t)
dt
The equilibrium points for (8) is ( ω , 0)and(0, 0).
µ

where 0 < S , I , R < 1. Because R = 1 ↑ (S + I), it is su"cient to .

retain only the first two equations. We are interested in the phase
portrait.
. 61 62
Disease-free equilibrium Disease-free equilibrium
It is worthwhile to linearise the nonlinear sytem (8) in order to be
able to use stability theory for linear system, from which the
eigenvalues of the Jacobian provides useful information regarding
the stability.
Set x1 = S ↑ S ε and x2 = I ↑ I ε where (S ε , I ε ) are the equilibrium
points.
Method 1: substitute x1 and x2 into (8) and ignore higher order
terms of the form x1 x2 .
Method 2: Expand f1 and f2 about the equilibrium points in the
equation below

dS
= f1 (S , I) = ↑ωS(t)I(t),
dt
dI
= f2 (S , I) = ωS(t)I(t) ↑ µI(t) (9)
dt

. 63 64
Disease-free equilibrium

65