Mathematical Modelling: Need, Techniques,

Classifications and Simple Illustrations

SIMPLE SITUATIONS REQUIRING MATHEMATICAL MODELLING
Consider the following problems:
(i) Find the height of a tower, say the Kutab Minar at New Delhi or the leaning
tower at Pisa (without climbing it!)
(ii) Find the width of a river or a canal (without crossing it!)
(iii) Find the mass of the Earth (without using a balance!)
(iv) Find the temperature at the surface or at the centre of the Sun (with¬ out
taking a thermometer there!)
(v) Estimate the yield of wheat in India from the standing crop (without cutting
and weighing the whole of it!)
(vi) Find the volume of blood inside the body of a person (without bleeding him to
death!)
(vii) Estimate the population of China in the year 2000 A.D. (without waiting till
then!)
(viii) Find the time it takes a satellite at a height of 10,000 kms, above the
Earth’s surface to completely one orbit (without sending such a satellite into
orbit!)
(ix) Find the effect on the economy of 30 per cent reduction in income-tax
(without actually reducing the rate!)
(x) Find the gun with the best performance when the performance depends on
ten parameters, each of which can take 10 values (without manufacturing 1010
guns!)
(xi) Estimate the average life span of a light bulb manufactured in a factory
(without lighting each bulb till it gets fused!)
(xii) Estimate the total amount of insurance claims a company has to pay next
year (without waiting till the end of that year!)
All these problems and thousands of similar problems can be and have been
solved through mathematical modelling.
One technique of solving the above problems is similar to that of solving ‘word
problems’ in algebra. Suppose the age of a father is four times the age of his son
and we are told that after five years, the age of the father will be only three
times the age of the son. We have to find their ages. Let x be the age of the
father and y be the age of the son, then the data of the problem gives
x = 4y, x + 5 = 3(y + 5), (1)
giving x = 40, y = 10. The two equations of (1) give a mathematical model of the
biological situation, so that the biological problem of ages is reduced to the
mathematical problem of the solution of a system of two algebraic equations.
The solution of the equations is finally interpreted biologically to give the ages of
the father and the son.
In the same way to solve a given physical, biological or social problem, we first
develop a mathematical model for it, then solve the model and finally interpret
the solution in terms of the original problem.
One principle of great importance to science is the following. Whenever we want
to find the value of an entity which cannot be measured directly, we introduce
symbols x, y, z,. . . to represent the entity and some others which vary with it,
then we appeal to laws of physics, chemistry, biology or economics and use
whatever information is available to us to get relations between these variables,
some of which can be measured or are known and others which cannot be
directly measured and have to be found out. We use the mathematical relations
developed to solve for the entities which can¬ not be measured directly in terms
of those entities whose values can be measured or are known.
The mathematical, relations we get may be in terms of algebraic, trans¬
cendental, differential, difference, integral, integro-differential, differential-
difference equations or even in terms of inequalities. Thus
For (i), we try to express the height of the tower in terms of some distances and
angles which can be measured on the ground,
For (ii), we try to express the width of the river in terms of some distances and
angles which can be measured on our side of the river,
For (iii), we try to express the mass of the Earth in terms of some known masses
and distances,
For (iv), we try to express the temperatures at the surface and the centre of the
Sun in terms of the properties of light received from its surface,
For (v), we try to find the area under wheat and the average yield per acre by
cutting and weighing the crop from some representative plots,
For (vi), we inject some glucose into the blood stream and find the increase in
the concentration of sugar in blood,
For (vii), we extrapolate from data from previous censuses or develop a model
expressing the population as a function of time,
For (viii), we try to use Newton’s laws to get a relation between the orbital period
and the height of the satellite above the surface of the Earth,
For (ix), we examine the effects of similar cuts in the past or develop a
mathematical model giving relation between income-tax cuts, purchasing power
in the hands of individuals and its effects on productivity and inflation etc.
For (x), we develop a theory of internal ballistics of guns based on laws of
burning of propellants, motion of gases inside gun and motion of the shot inside
it.
For (xi), we take a random sample of bulbs, find their life-span and use statistical
inference models to estimate the life span for the population of bulbs.
For (xii), we use probabilistic models for life expectancy of individuals.

THE TECHNIQUE OF MATHEMATICAL MODELLING

Mathematical modelling essentially consists of translating real world problems
into mathematical problems, solving the mathematical problems and interpreting
these solutions in the language of the real world (Figure 1.4).

This is expressed figuratively by saying that we catch hold of the real world
problem in our teeth, dive into the mathematical ocean, swim there for some
time and we come out to the surface with the solution of the real world problem
with us. Alternatively we may say that we soar high into the mathematical
atmosphere along with the problem, fly there for some time and come down to
the earth with the solution.
A real world problem, in all its generality can seldom be translated into a
mathematical problem and even if it can be so translated, it may not be possible
to solve the resulting mathematical problem. As such it is quite often necessary
to ‘idealise’ or ‘simplify’ the problem or approximate it by another problem which
is quite close to the original problem and yet it can be translated and solved
mathematically. In this idealisation, we try to retain all the essential features of
the problem, giving up those features which are not very essential or relevant to
the situation we are investigating.
Sometimes the idealisation assumptions may look quite drastic. Thus for
considering the motions of planets, we may consider the planets and Sun as
point masses and neglect their sizes and structures. Similarly for considering the
motion of a fluid, we may treat it as a continuous medium and neglect its
discreate nature in terms of its molecular structure. The justification for such
assumptions is often to be found in terms of the choseness of the agreement
between observations and predictions of the mathematical models.
This leads us to modify Figure 1.4 to the following Figure 1.5.
If the comparison is not satisfactory, we modify either the idealisation
assumptions or search for another structure for the mathematical model.
This leads to the following twelve-point procedure for solving problems through
mathematical modelling:
(i) Be clear about the real world situation to be investigated. Find all its essential
characteristics relevant to the situation and find these aspects which are
irrelevant or whose relevance is minimal. It is important to decide what aspects
must be considered and what aspects can be ignored.
(ii) Think about all the physical, chemical, biological, social, economic laws that
may be relevant to the situation. If necessary collect some data and analyse it to
get some initial insight into this situation.
(iii) Formulate the problem in Problem Language (PL).
(iv) Think about all the variables X1, X2, . . ., Xn and parameters
a1, a2, ….., am
involved. Classify these into known and unknown ones.
(v)Think of the most appropriate mathematical model and translate the problem
suitably into mathematical language (ML) in the form

(
f j x i , ah ,
∂
∂ xi )
,∫ … .. d x i ,d ≤0 (2)

i.e. in terms of algebraic, transcendental, differential, difference, integral,

integro-differential, differential-difference equations or inequations.
(vi) Think of all possible ways of solving the equations of the model. The methods
may be analytical, numerical or simulation. Try to get as far as possible
analytically, supplement this with numerical and computer methods when
necessary and use simulation when warranted.
(vii) If a reasonable change in the assumptions makes analytical solution
possible, investigate the possibility. If new methods are required to solve the
equations of the model, try to develop these methods.
(viii) Make an error analysis of the method used. If the error is not within
acceptable limits, change the method of solution.
(ix) Translate the final solution into P.L.
(x) Compare the predictions with available observation or data. If agree¬ ment is
good, accept the model. If the agreement is not good, examine the assumptions
and approximations and change them in the light of the discrepancies observed
and proceed as before.
(xi) Continue the process till a satisfactory model is obtained which explains all
earlier data and observations.
(xii) Deduce conclusions from your model and test these conclusions against
earlier data and additional data that may be collected and see if the agreement
still continues to be good.
This technique will be repeatedly used in the book. However, since most of the
models considered are already well-established models, their validation will not
be considered, but for modelling of new situations, validation is essential.
CLASSIFICATION OF MATHEMATICAL MODELS
(a) Mathematical Models (M.M.) may be classified according to the subject matter
of the models. Thus we have M.M. in Physics (Mathematical Physics) M.M. in
Chemistry (Theoretical Chemistry); M.M. in Biology (Mathematical Biology), M.M.
in Medicine (Mathematical Medicine), M.M. in Economics (Mathematical
Economics and Econometrics), M.M. in Psychology (Mathematical Psychology),
M.M. in Sociology (Mathematical Sociology), M.M. in Engineering (Mathematical
Engineering) and so on.
We have similarly M.M. of transportation, of urban and regional plan¬ ning, of
pollution, of environment, of oceanography, of blood flows, of genetics, of water
resources, of optimal utilization of exhaustible and renew¬ able resources, of
political systems, of land distribution, of linguistics and so on.
In fact every branch of knowledge has two aspects, one of which is theoretical,
mathematical, statistical and computer-based and the other of which is
empirical, experimental and observational. Mathematical Modelling is essential
to the first of these two aspects.
We have separate books on mathematical models in each of the areas we have
mentioned above and in many others. One can spend a life-time specialising in
mathematical models in one specified area alone.
(b) We may also classify Mathematical Models according to the mathematical
techniques used in solving them. Thus we have Mathematical Modelling (M.M.)
through classical algebra, M.M. through linear algebra and matrices, M.M.
through ordinary and partial differential equations, M.M. through ordinary and
partial difference equation, M.M. through integral equations, M.M. through
integro-differential equations, M.M. through differential-difference equations,
M.M. through functional equa¬ tions, M.M. through graphs, M.M. through
mathematical programming, M.M. through calculus of variations, M.M. through
maximum principle and so on.
Again there are books on each of these techniques. However in most of these
books, most of the space is devoted to explaining the theory of the technique
concerned and applications are given as illustrations only. Mathematical
modelling aspect is seldom emphasised.
In books of category (a), mathematical modelling is emphasised and techniques
are considered of secondary importance (through this is not always the case) but
the models belong to one specified field of knowledge. In books of category (b)
the theory of the technique is emphasised and readymade models are used to
illustrate the technique. In the present book, we assume the knowledge of the
basic theory of each technique and lay emphasis mainly on mathematical
modelling and applications of the technique. In particular we consider when
models in terms of specific techniques may be relevant. Books of category (a)
consider applications of mathematics in one specified field of knowledge, but use
a diversity of mathematical techniques. Books of category (b) use a single
technique, but consider application in a diversity of fields of knowledge. In the
present book, we consider both a diversity of techniques and a diversity of fields
of knowledge.
(c) Mathematical Models may also be classified according to the purpose we
have for the model. Thus we have Mathematical Models (M.M.) for Description,
M.M. for Insight, M.M. for Prediction, M.M. for Optimization, M.M. for Control and
M.M. for Action.
(d) Mathematical Models may also be classified according to their nature. Thus
(i) Mathematical Models may be Linear or Non-Linear according as the basic
equations describing them are linear or non-linear.
(ii) Mathematical Models may be Static or Dynamic according as the time-
variations in the system are not or are taken into account.
(iii) Mathematical Models may be Deterministic or Stochastic according as the
chance factors [Link] or are taken into account.
(iv) Mathematical Models may be Discrete or Continuous according as the
variables involved are discrete or continuous.
Linear, static and deterministic models are usually easier to handle than non-
linear, dynamic and stochastic models and in general in any discipline, these are
the first to be considered.

Continuous-variate models appear to be easier to handle than the discrete-

variate models, due to the development of calculus and differential equations. In
fact in many disciplines, these were developed first. However continuous models
are simpler only when analytical solutions are available, otherwise we have to
approximate a continuous, model also by a discrete model so that these can be
handled numerically.
There are of course models which involve both discrete and continuous variates
simultaneously.
Essentially most realistic models are non-linear, dynamic and stochastic. We use
linear, static or deterministic models because these are easier to handle and give
good approximate answers to our problems.
When the variables are essentially discrete, we may still use continuous models
to be able to use calculus and differential equations. Similarly when the variables
are essentially continuous, we may still use a discrete model to be able to use
computers.
SOME CHARACTERISTICS OF MATHEMATICAL MODELS
(i) Realism of models'. We want a mathematical model to be as realistic as
possible and to represent reality as closely as possible. However, if a model is
very realistic, it may not be mathematically tractable. In making a mathematical
model, there has to be a trade-off between tractability and reality.
(ii) Hierarchy of models'. Mathematical modelling is not a one-shot affair. Models
are constantly improved to make them more realistic. Thus for every situation,
we get a hierarchy of models, each more realistic than the preceding and each
likely to be followed by a better one.

(iii) Relative precision of models: Different models differ in their preci¬ sion and
their agreement with observations.
(iv) Robustness of models: A mathematical model is said to be robust if small
changes in the parameters lead to small changes in the behaviour of the model.
The decision is made by using sensitivity analysis for the models.
(v) Self-consistency of models : A mathematical model involves equations and
inequations and these must be consistent, e.g. a model cannot have both x + y
> a and x+y < a. Sometimes the inconsistency results from inconsistency of
basic assumptions. Since mathematical inconsistency is relatively easier to find
out, this gives a method of finding inconsistency in requirements which social or
biological scientists may require of their models. A well-known example of this is
provided by Arrow’s Impossibility Theorem.
(vi) Oversimplified and overambitious models: It has been said that mathematics
that is certain does not refer to reality and mathematics that refers to reality is
not certain. A model may not represent reality because it is oversimplified. A
model may also be overambitious in the sense that it may involve too many
complications and may give results accurate to ten decimal places whereas the
observations may be correct to two decimal places only.
(vii) Complexity of models: This can be increased by subdividing vari¬ ables, by
taking more variables and by considering more details. Increase of complexity
need not always lead to increase of insight as after a stage, diminishing returns
begin to set in. The art of mathematical modelling consists in stopping before
this stage.
(viii) Models can lead to new experiments, new concepts and new mathe¬
matics: Comparison of predictions with observations reveals the need for new
experiments to collect needed data. Mathematical models can also lead to
development of new concepts. If known mathematical techniques are not
adequate to deduce results from the mathematical model, new mathematical
techniques have to be developed.
(iv) A model may be good, adequate, similar to reality for one purpose and not
for another: Thus we may need different models for explaining different aspects
of the same situation or even for different ranges of the variables. Of course in
this case, search for a unified model continues.
(x) Models may lead to expected or unexpected predictions or even to non¬
sense: Usually models give predictions expected on common sense consider¬
ations, but the model predictions are more quantitative in nature. Sometimes
they give unexpected predictions and then they may lead to break-throughs or
deep thinking about assumptions. Sometimes models give prediction completely
at variance with observations and then these models have to be drastically
revised.
(xi) A model is not good or bad; it does or does not fit: Models may lead to nice
and elegant mathematical results, but only those models are accept¬ able which
can explain, predict or control situations. A model may also fit one situation very
well and may give a hopeless fit for another situation.
(xii) Modelling forces us to think clearly. Before making a mathematical model,
one has to be clear about the structure and essentials of the situation.
(xiii) Sticking to one model may prevent insight: A model helps thinking, but it
can also direct thinking in one narrow channel only. Sometimes insight is
obtained by breaking with traditional models and designing entirely new ones
with new concepts.
(xiv) Inadequate models are also useful'. Since they lead us to search for aspects
which may have been neglected at first. Failure can be prelude to successes if we
can find the reasons for these failures.
(xv) Non-feedback models are improper'. A model must include the possi¬ bility
of its improvement in the light of the experimental or observational data.
(xvi) Partial modelling for subsystems'. Before making a model for the whole
system, it may be convenient to make partial models for subsystems, test their
validity and then integrate these partial models into a complete model.
Sometimes existing models are combined to give models for bigger systems.
Often models are unified so that the general model includes the earlier models
as special cases.
(xvii) Modelling in terms of modules'. One may think of models for small modules
and by combining them in different ways, one may get models for a large
number of systems.
(xviii) Imperfections of models and cost of modelling'. No model is perfect and
every model can be improved. However each such improvement may cost time
and money. The improvement in the model must justify the invest¬ ment made
in this process.
(xix) State variables and relations: For making a mathematical model, one has
first to identify the state variables and then specify the relations between them.
The right choice of state variables is of the utmost importance.
(xx) Estimation ofparameters: Every model contains some parameters and these
have to be estimated. The model must itself suggest experiments or
observations and the method of calculation of these parameters. Without this
explicit specification, the model is incomplete.
(xxi) Validation by independent data: Sometimes parameters are estimated with
the help of same data and the same data are used to validate the model. This is
illegitimate. Independent data should be used to validate the model.
(xxii) New models to simplify existing complicated models: We start with simple
models, introduce more and more variables and more and more functions to
make the models more realistic and more complicated and with the additional
insights obtained, we should again be able to simplify the complex models.
(xxiii) Modelling => Mathematics + Discipline: For making a mathematical
models of a situation, one must know both mathematics and the discipline in
which the situation arises. Efforts to make a mathematical model without deeply
understanding the discipline concerned may lead to infructous models. Discipline
insight must both precede and follow mathematical modelling.
(xxiv) Transferability of mathematical models: A mathematical model for one
field may be equally valid for another field and may be validly transferred to
another field, but great care must be exercised in this process. A model which is
transferable to a number of fields is very useful, but no model should be thrust
on a field unless it is really applicable there.
(xxv) Prediction-validation-iteration cycle'. A mathematical model predicts
conclusions which are then compared with observations. Usually there is some
discrepancy. To remove this discrepancy, we improve the model, again predict
and again try to validate £nd this iteration is repeated till a satisfactory model is
obtained.
(xxvi) Models for strategic and tactical thinking: Models may be construc¬ ted for
determining guidelines for particular situations or they may be for determining
an overall strategy applicable to a variety of situations.
(xxvii) Constraints of additivity and normality. Models which are linear, additive
and in which the probability distribution follows the normal law are relatively
simpler, but relatively more realistic models have to be free from these
constraints.
(xxviii) Mathematical modellings and mathematical techniques: Emphasis in
applied mathematics has very often been on mathematical techniques, but the
heart of applied mathematics is mathematical modelling.
(xxix) Mathematical modelling gives new ideology and unity to applied
mathematics'. Thus operations research and fluid dynamics differ in their subject
matter as well as in techniques, but mathematical modelling is common to both.
(xxx) Non-uniqueness of models: A situation need not have only one
mathematical model and the existence of one model for it should not inhibit
search for better and different models.
(xxxi) Dictionary of mathematical models'. It is unlikely that we shall ever have a
complete dictionary of mathematical models so that our task will be only to
choose an appropriate model for a given situation. Familiarity with existing
models will always be useful, but new situations will always demand construction
of new models.
(xxxii) No prefabrication of models: Some pure mathematicians believe that
every consistent logical structure will one day model some physical situation.
This is likely to be an exception rather than the rule. There will always be a very
large number of mathematical structures without corres¬ ponding physical
models and there will always be physical situations without good mathematical
models. Search has to go on in both directions. Mathe¬ matics for modelling has
to be mainly motivated by the world around us.
(xxxiii) Mathematical modelling is an Art: It requires experience, insight and
understanding. Teaching this art is also another art.
(xxxiv) Criteria for successful models: These include good agreement between
predictions and observations, of drawing further valid conclusions, simplicity of
the model and its precision.
(xxxv) Generality and applicability of models: Laplace equation model applies to
gravitational potential, electro-static potential, irrotational flows and a variety of
other situations. There are some models applicable to a wide variety of
situations, while there are others which are applicable to specific situations only.
(xxxvi) Unity of disciplines through mathematical modelling: When a number of
different situations are represented by the same mathematical model, it reveals
a certain identity of structures of these situations. It can lead to a certain
economy of efforts and it can reveal a certain underlying unity between different
disciplines.

MATHEMATICAL MODELLING THROUGH GEOMETRY

(a)One of the earliest examples of mathematical modelling was that of
mathematical description of the paths of planets. Looked at from the Earth the
paths were not simple curves like circles or ellipses. The next curve known in
order of complexity was an epicycloid which is the locus of a point on a circle
which rolls on another fixed circle. The path of a planet was not even an
epicycloid. However, it was found possible to combine suitably a number of these
epicycloidal curves or epicycles to describe the paths of all the planets. This was
highly successful, though quite a complicated model.
(b)Another geometric modelling mirrors for burning enemy ships by
(c) A similar geometric modelling is involved in constructing an elliptic sound
gallery so that the sound produced at one focus can be heard at the other focus
after being reflect¬ ed back from every point of the ellipse (Figure 1.7).
(d) Based on the observations of Copernicus, Kepler showed that each planet
moves in an ellipse with the Sun at one focus. Thus the heliocentric theory of
planetory motion completely simplified the description of the paths of the
planets. The earlier geocentric theory required complicated combination of
epicycloids. Both the models are correct, but the heliocen¬ tric model is much
simpler than the geocentric model. However both the models were models for
description only. Later Newton showed that the elliptical orbit followed from the
universal law of gravitation and thus this model became a model for
understanding. Still later in 1957, the elliptic orbits were used as orbits of
satellites. At this stage, the model became a model for control. Now the same
model can be used for getting optimal orbits for the satellites and as such it can
also be used as a model for optimization.

(e) A fifth geometrical model is involved in the use of Fermat’s principle of least
time which states that light travels from one point to another in such a way as to
take least possible time. One immediate consequence of this is that in a
homogeneous medium, light travels in a straight line, since a straight line
corresponds to the shortest distance between two points.
If however light travels from point A to point B after being reflected from a mirror
CD, the light ray will be incident at such a point O of the mirror (Figure 1.8) so
that AO + OB is minimum or such that AO + OB' is minimum, where B' is the
mirror- image of B in CD. But TO + OB' is minimum when AOB' is a straight line.
This gives
Geometric Principle

∠ AOY =∠ BOY orangle of incidence=angle of reflection ¿ (3)

MATHEMATICAL MODELLING THROUGH ALGEBRA
(a) Finding the Radius of the Earth
This model was used about two thousand years ago. A and B are two points on
the surface of the Earth with the same longitude and d miles apart. When the
Sun is vertically above A (i.e. it is in the direction OA, where O is the centre of
the Earth) (Figure 1.9), the Sun’s rays make an angle of θ° with the vertical at B
(i.e. with the line OB). If a miles is the radius of the Earth, it is easily seen from
Figure 1.9 that

d θ 360 d
= or a= ¿ (4)
2 πa 360 2 πθ

(b) Motion of Planets

The orbit of each planet is an ellipse with the Sun at one focus. However, the
ellipticities of the orbits are very small, so that as a first approximation, we can
take these orbits as circles with Sun at the centre. Also we know that the planets
move under gravitational attraction of the Sun and that for motion in a circle with
uniform speed v, a central acceleration v 2 /r is required.

If the masses of the Sun and the planet are S and P respectively, we get

2
GPS P v 2 GS
2
= or v = , ¿ (5)
r r r
where G is the constant of gravitation. Further if T is the periodic time of the
planet, we have

vT =2 πr ¿ (6)
Eliminating v between (5) and (6), we get
2 3
2 4π r
T = ¿ (7)
GS
If T₁, T₂ are the time periods of two planets with orbital radii r₁, r₂, then
2 2 3 3
T 1 /T 2=r 1 /r 2 , ¿ (8)
so that the squares of the periodic times are proportional to the cubes of the
radii of the orbits.

(c) Motions of Satellites

Satellites move under the attraction of the Earth in the same way as the planets
move under the attraction of the Sun, so that we get

¿
where E is the mass of the Earth, a is the radius of the Earth and h₁, h₂ are the
heights of the satellites above the Earth’s surface. Also if g is the acceleration
due to gravity at the Earth’s surface, then

GmE 2
mg= 2
or≥¿ g a ¿ (10)
a
From (9) and (10)

¿
(d)
We can solve the problem of 1.5 (e) by using algebraic method. In Figure 1.8,

¿
Simplifying
2 2 2 2
2 c k 2 k
4 x (1− 2 )−4 cx (1+ 2 )+ 4 a −(m+ ) =0 ;k 2=a2−b2−c2 ¿(13)
m m m
Since x is real, we get
2 2 2 2 2
2 k c 2 2 k
c (1+ 2
) ≥(1− 2 )(4 a −m (1+ 2 ) )
m m m
or

¿
From Figure 1.8, the second factor is positive. As such (14) gives

¿
Thus the minimum value of m is {¿and when m has this value, the two roots of
the quadratic (13) are equal and each is given by
2
k
4 c (1+ 2
)
m 2 ac
2 x= 2
or . ¿ (16)so that
c a+b
4(1− 2 )
m
x c−x
= or α =β . ¿ (17)
a b
(e)
By using the algebraic result that the arithmetic mean of n positive numbers ≥
the geometric mean of these numbers and the equality sign holds iff the
numbers are equal, we can deduce that:
(i) If the sum of n positive numbers is constant, then their product is maximum
when the numbers are equal.
(ii) If the product of n positive numbers is constant, then their sum is minimum
when the numbers are equal.
(iii) Of all rectangles with a given perimeter, the square has the maximum area.
(iv) Of all rectangles with a given area, the square has the minimum perimeter.
(v) Of all rectangular parallelepipeds with a given perimeter, the cube has the
maximum volume.
(vi) Of all rectangular parallelepipeds with a given volume, the cube has the
minimum perimeter.
(vii) The quantity x p ¿is maximum when x / p=(a−x)/q .

(f)
In the same way, we have
w1 w2 wn
w 1 x 1 +w 2 x 2+ …+w n x n ≥ x 1 x 2 … x n , ¿ (18)

where w 1 , w 2 ,… , w nare positive weights with

n

∑ ❑ wi=1
i=1

and the equality sign holds iff x 1=x 2=…=x n.

Now let ( p1 , p 2 , … , p n) , (q 1 , q 2 , … , q n) be two probability distributions such that

n n

∑ pi =1 , ∑ qi=1 , p i> 0 , qi >0. ¿ (19)

i=1 i=1

Putting w i=q i, x i= pi /q iin (18), we get

( )
n qi
pi n pi
∑ qi ≥ ∑ qi
q i i=1 qi
i=1

or
n n
pi
∑ qi ln ⁡ q ≤ ln ⁡∑ p i=0 , ¿ (20)
i=1 i i=1

so that
n n
qi p
∑ qi ln ⁡ p ≥ 0 , ∑ pi ln ⁡ q i ≥ 0. ¿ (21)
i=1 i i=1 i

and the equality sign holds iff pi=qifor all i. The inequality (21) is known as
Shannon’s inequality and is useful in the development of maximum-entropy
models.

(g)

1
Putting q i= in (21), we get
n
n n n

∑ pi ln ⁡pi +∑ pi ln ⁡n ≥ 0 or ∑ pi ln ⁡p i ≥−ln ⁡n
i=1 i=1 i=1

or
n n
1 1
−∑ pi ln ⁡pi ≤ ln ⁡n=−∑ ln ⁡ ¿ (22)
i=1 i=1 n n
The expression
n
S=−∑ p i ln ⁡p i
i=1

is called the entropy of the probability distribution ( p1 , p 2 , … , p n) , and (22) shows

that its maximum value is ln n , and this maximum value is attained when all the
probabilities are equal.
This result is of great importance in information theory, coding theory, and in
the development of maximum-entropy models.

MATHEMATICAL MODELLING THROUGH TRIGONOMETRY

(a) Finding the Distance of the Moon
From two points A, B on the surface of the Earth, with the same longitude—one in
the Northern hemisphere and the other in the Southern hemisphere—measure
angles θ₁, θ₂ between verticals at A and B and the directions of the centre of the
Moon (Figure 1.11).

If d is the distance of the centre of the Moon’s disc from the centre of Earth,
Figure 1.11 gives

d a d a
= , = ¿ (23)
sin θ1 sin ⁡(θ1−ψ 1 ) sinθ 2 sin ⁡(θ2−ψ 2 )

Also

ψ 1 +ψ 2=α =φ 1+ φ2 ¿ (24)

where φ 1 is the Northern latitude of A and φ 2 is the Southern latitude of B. Since

φ 1, φ 2 are known, ψ 1 +ψ 2 is known. Eliminating ifn, ψ 1 ,ψ 2 from (23) and (24), we
get d in terms of a, θ1 , θ2 which are all known.

(b) Finding the Distance of a Star

For a star, the base line provided by AB is too small and we choose the largest
distance available to us viz. the positions of the Earth six months apart. We can
measure angles SAX and SBX and since we know length AB which is about 186
million miles, we can determine AX and BX from triangle ABX (Figure 1.12).
(c) Finding Length of the Day
The length of the day depends on the latitude of the place and declination of the
Sun on the particular day.
In astronomy, we are concerned both with directions and distances of heavenly
bodies. In spherical astronomy, we are concerned only with their directions from
the observer.
We draw a sphere with unit radius with the observer at its centre. This sphere will
be called the celestial sphere. If we join the observer to any heavenly body, this
line will meet the celestial sphere on some point which will be called the position
of the heavenly body on the celestial sphere.
If A, B, C are positions of three heavenly bodies on the celestial sphere and we
join these by great circle arcs, we get a spherical triangle ABC, the lengths of
whose sides are measured by the angles subtended by these arcs at the centre.
The relations between sides and angles of a spherical triangles are given by

sin A sin B sin C

= =
sin a sin b sin c
and
cos a = cos b cos c + sin b sin c cos A (26)
The point vertically above the observer is called his zenith and is denoted by Z.
Thus OZ gives the direction of the line joining the centre of the Earth to the
observer. Similarly the direction of the axis of rotation of the Earth determines a
point P on the celestial sphere and the arc PZ is equal to the angle between OP
and OZ and is thus equal to π /2 — φ , where φ is the latitude of the observer.

Due to the daily rotation of the Earth about its axis OP, the Sun
appears to move in a plane perpendicular to OP (Figure 1.13). The Sun
rises above the horizon at A, reaches its highest position at B, sets at the point C
and remains invisible during its motion on arc CDA.
The length of the day is determined by the movement of the Sun across
the sky from A to B and from B to C. This movement is determined by
the angle H, measured in radians. One hour corresponds to a movement
π
of radians.
12
Let us now consider the spherical triangle ZPA, where:

π π
 ZP= −ϕ , ZA=
2 2
 Z is the pole of the horizon
 ϕ is the latitude of the observer
π
 PA= −δ , where δ is the declination of the Sun
2
Using equation (26) in the spherical triangle ZPA:

co s ( π2 )=co s( π2 −φ ) ⋅co s( π2 −δ )+ si n( π2 −φ ) ⋅si n( π2 −δ ) ⋅cos ⁡H

This simplifies to:

cos ⁡H=−tan ⁡φ ⋅ta n δ ( 27)

If φ and δ are known, then equation (27) determines H, and the length of the
day:

24 H
¿2 H x =24
2π π
Solar Declination Values Throughout the Year
∘ ∘
 δ=−23.5 on 23rd December, δ=0 on 21st March
∘ ∘
 δ=23.5 on 23st June, δ=0 on 21st September
In the morning, twilight starts when the Sun is at point E, where:
∘ '
 ∠ ZE=108 , ∠ ZPE=H
Using triangle ZPE and equation (26):
∘ '
cos ⁡108 =sinφ ⁡sin ⁡δ+cosφ ⁡cos ⁡δ cos ⁡H ¿(28)
From equations (27) and (28), we can find H and H', and therefore:
'
H −H =total length of twilight in morning or evening
So, the total twilight duration is: 24 (H ' −H )π hours

MATHEMATICAL MODELLING THROUGH CALCULUS

Consider the model of Section 1.5(e) and 1.6(d). The distance travelled by light
from A to B is

¿
so that

¿
¿
So that m is minimum when, using figure 1.8,

x c−x
=
√ a + x √ b2 +¿ ¿ ¿
2 2

(b) Law of Refraction of Light

Consider the problem of refraction of light from a point A in vacuum to a point B
(Figure 1.14) in a medium of refractive index μ. If light travels with velocity V in
vacuum, it travels with velocity V / μin the medium so that the time T of travel is
given by:

¿
Then,

¿
¿
Thus, T is minimum when:
 x (c−x)
=μ ⋅ orsin ⁡α=μ sin ⁡β ¿ (36)
√ a2 + x 2
√ b2 +¿ ¿ ¿
(c) EOQ Model for Inventory Control

Let the total demand for a commodity be D units in a year and let orders of q units
be placed D/q times in a year. Let the cost of ordering be C 1per order so that the
total cost of ordering is:

D
C1⋅
q
In each order interval of duration q / D, the quantity in stock falls uniformly from q
to 0, so the average stock in this period is q /2. Let the cost of storing be C 2per
unit per unit time, so that the total storage cost is:

q D q
C 2 ⋅ ⋅ =C 2 ⋅
2 q 2
Thus, the total cost of ordering and storing is:

D q
f (q)=C 1 ⋅ +C 2 ⋅ ¿ (37)
q 2
Differentiating:

' D 1 '' 2 C1 D
f (q)=−C 1 ⋅ 2
+C 2 ⋅ , f (q)= 3
>0 ¿ (38)
q 2 q
So the total cost is minimum when:

q=
√ 2C 1 D D
C2
, =
q
D C2
2C 1
¿ (39)
√
(d) Triangle of Given Perimeter with Maximum Area

The square of the area of a triangle with semi-perimeter sis:

2 4 3 2
Δ =s(s−a)(s−b)(s−c)=s −s (a+b+ c)+ s (ab+bc +ca)−sabc
2
4 2 s 2 2 2 2
¿ s −2 s + [4 s −a −b −c ]−sabc
2
2
4 s 2 2 2
¿s − ( a + b +c )−sabc ¿ (40)
2
The constraint to be maximized is:

2 s=a+b+ c ¿ (41)
Using Lagrange's method:
2 2 2
s a+sbc s b + sca s c + sab
= =
1 1 1
 2 2 2
s a +abc s b + abc s c +abc
= =
a b c
2 2 2 2
s (a +b ) s (b + c )
=
a−b b−c
Or a+b=b+c or a=c
Similarly, b = c so that a = b = c
(e) Parallelepiped with Given Perimeter and Maximum Volume

We have to maximize xyz subject to x + y + z=3 a

The Lagrangian is:

L=xyz −λ(x + y + z−3 a) ¿ (44)

This is maximum when:

yz=zx =xy or x= y=z ¿ (45)

Thus, the parallelepiped of maximum volume is a cube. The parallelepiped of
minimum volume has of course zero volume.
Alternatively,

V =xy ( 3 a−x− y ) ,
∂V 2 ∂V 2
=3 ay−2 xy− y , =3 ax−x −2 xy ¿(46)
∂x ∂y
1. Volume Maximization via Second Partial Derivatives
2 2 2
∂V ∂ V ∂ V
2
=−2 a , 2
=−2 a , =3 a−2 x−2 y ¿ (47)
∂x ∂y ∂x∂y
Putting the first derivatives equal to zero, we get x= y =a, and at this point:
2 2
∂V ∂V
2
=−2 a<0 , 2 =−2 a< 0 ,
∂x ∂y
2 2 2 2
∂V ∂V ∂ V 2
2
⋅ 2 −( ) =a > 0 ¿ (48)
∂x ∂ y ∂ x ∂ y
So the volume is maximum when x= y =z=a , i.e., when the rectangular
parallelepiped is a cube.
(f) Mathematics of Business

(i) Let the revenue obtained by selling a commodity at price pbe given by:
2
R( p)=ap−b p ¿ (49)
So that:
' ''
R ( p)=a−2 bp , R ( p)=−2 b ¿ (50)
 a
Thus, the profit is maximum when the price is .
2b
(ii) The revenue on selling x items is R(x ), and the cost of x items is C (x). Then
the profit function P(x )is given by:

P(x )=R(x )−C (x) ¿ (51)

So, that
' ' ' '' '' ''
P (x)=R ( x)−C (x), P (x )=R ( x)−C (x ) ¿ (52)
Thus, the profit is maximum when marginal revenue (revenue from selling one
additional item) is equal to marginal cost (cost of producing one additional
item).

(iii) If the total cost of producing q items is a+ bq+c q 2, then the average cost per
item is:

a
φ (q)= +b +cq ¿ (53)
q
' −a '' 2a
So that:φ (q)= 2
+c , φ (q)= 3 ¿ (54)
q q

The average cost per item is minimum when: q=

√ a
c
(iv) If the profit in a district for x sales representatives is:
2
f ( x)=−a x + bx−c ¿ (55)

Then: f ' ( x)=−2 ax +b , f ' ' ( x )=−2 a ¿ (56)

b
Thus, the profit is maximum for: x=
2a
(v) If the number of travelling passengers in a city bus system is a – bp, where, p
is the price of a ticket, then the total revenue is ap – bp 2 and this is maximum
when the price is a /2b .

(vi) The demands for two related products sold by a company are given by:

q 1=a1 −a2 p 1+ a3 p 2 and q2 =b1−b 2 p1 +b 2 p2 ¿ (57)

 Ans the total sales value is:
2 2
S= p 1 q 1+ p 2 q 2=a 1 p1 +b1 p 2−a 2 p1 +b2 p1 p2 −b2 p 2 ¿ (58)
 To maximize the sale value, we place zero derivatives:

a 1−2 a2 p 1−(a3 +b2 ) p2=0 ¿

a 1−2 b3 p 2−(a3 +b2 ) p1=0 ¿ (59)

LIMITATIONS OF MATHEMATICAL MODELLING
There are thousands of mathematical models which have been successfully
developed and applied to get insight into tens of thousands of situations. In fact
mathematical physics, mathematical economics, operations research,
biomathematics etc. are almost synonymous with mathematical modelling.
However there are still an equally large or even a larger number of situations
which have not yet been mathematically modelled either because the situations
are sufficiently complex or because mathematical models formed are
mathematically intractable.
The development of powerful computers has enabled a much larger number of
situations to be mathematically modelled. Moreover it has been possible to make
more realistic models and to obtain better agreement with observations.
However, successful guidelines are not available for choosing the number of
parameter and of estimating the values for these parameters. In fact reasonably
accurate models can be developed to fit any data by choosing number of
parameters to be even five or six. We want a minimal number of parameters and
we want to be able to estimate them accurately.
Mathematical Modelling of large-scale systems presents its own special
problems. These arise in study of world models and in global models of
environment, oceanography, economic conditions, pollution control etc.
However mathematical modellers from all disciplines-mathematics, statis¬ tics,
computer science, physics, engineering, social sciences—are meeting the
challenges with courage. Six international conferences on Mathematical
Modelling have been held and a large number of specialised conferences on
mathematical modelling have been organised. Teaching of Mathematical
Modelling has not been neglected and the first three international conference on
the Teaching of Mathematical Modelling have already been held.

Advanced modeling applications to water

resources
1) Flood forecasting & inundation mapping (real-time + planning)
What/Why: Predict river flows and flood extents for warnings, design, and
zoning.
Core models:
 Rainfall-runoff (HBV/GR4J/SWAT), routing (Muskingum, kinematic/dynamic
wave).
 1D/2D shallow-water (Saint-Venant) for floodplain hydraulics.
Key equations: Continuity + momentum; Manning/Chézy friction.
Data: Radar/satellite/gauge rainfall, DEMs, channel geometry, roughness,
levees.
Advanced: Ensemble meteorology, data assimilation (EnKF) to update
states, 2D unstructured meshes, GPU solvers, ML surrogates for rapid
what-ifs.
Outputs: Hydrographs, depth/velocity grids, hazard maps, AEP curves.
Pitfalls: DEM artifacts, roughness–slope equifinality, ignoring
bridge/culvert losses.
2) Drought & low-flow risk modeling
What/Why: Assess reliability of water supply and environmental flows under
deficits.
Core models: Stochastic streamflow generation (ARIMA, block bootstrapping),
rainfall-runoff with soil moisture states.
Advanced: Non-stationary frequency analysis, multi-year hydrologic state
persistence, Bayesian updating, climate conditioned ensembles.
Outputs: Reliability, resilience, vulnerability; reservoir rule curves.
Pitfalls: Assuming stationarity; using annual indices for sub-seasonal
operations.
3) Integrated basin models (surface–groundwater coupling)
What/Why: Quantify exchanges between rivers, aquifers, irrigation, wetlands.
Core models: MIKE SHE / GSFLOW (PRMS + MODFLOW), coupling overland,
unsaturated (Richards), and saturated flow (Darcy).
Data: Soil maps, hydraulic conductivities, canal/diversion schedules, ET (RS).
Advanced: Iterative coupling schemes, calibration with PEST/iTOUGH, multi-
objective metrics (flows + heads), assimilation of GRACE/SMAP.
Outputs: Baseflow fractions, recharge maps, depletion by pumping.
Pitfalls: Over-parameterization; poor representation of irrigation return flows.
4) Groundwater development & drawdown management
What/Why: Design wellfields, prevent interference and subsidence.
Core models: MODFLOW (confined/unconfined), well losses, stream–aquifer
packages.
∂h
Math: S s =∇ ⋅(K ∇ h)+W .
∂t
Advanced: Surrogate optimization for pumping (NSGA-II), uncertainty bands
via MCMC, travel time capture zones (MODPATH).
Outputs: Drawdown cones, sustainable yields, capture fractions.
Pitfalls: Calibrating to heads only (add flows/stream gains), ignoring aquitards.
5) Contaminant fate & transport (nitrate, arsenic, salinity, PFAS)
What/Why: Predict plume migration and remediation timelines.
Core: Advection–dispersion–reaction in porous media (MT3DMS/RT3D, SEAWAT
for density).
Math: ∂ C /∂ t+ v ⋅ ∇ C=∇ ⋅(D ∇ C)−kC+ S .
Advanced: Reactive networks (Monod, dual-porosity), geostatistical
heterogeneity, inverse modeling for source identification, polynomial chaos for
uncertainty.
Outputs: Concentration plumes, mass discharge, compliance time.
Pitfalls: Dispersion vs. heterogeneity confusion; grid Peclet violations.
6) River/reservoir water-quality & eutrophication
What/Why: Assess oxygen, nutrients, algae; meet TMDLs.
Tools: QUAL2K/CE-QUAL-W2/WASP with temperature, DO–BOD, N–P cycles,
algae.
Advanced: Light/temperature limitation, benthic fluxes, sediment diagenesis,
load-reduction scenarios with Pareto trade-offs (cost vs WQ).
Outputs: Longitudinal profiles, exceedance probabilities, critical loads.
Pitfalls: Using steady flows for flashy rivers; uncalibrated reaeration.
7) Sediment transport & river morphodynamics
What/Why: Predict aggradation, scour, navigation depth, habitat.
Core: Saint-Venant + Exner (bed mass balance); bedload/suspended formulas
(Meyer-Peter–Müller, Engelund–Hansen).
Advanced: 2D/3D morphodynamic models (Delft3D/TELEMAC), bank erosion
modules, morphological acceleration factors.
Outputs: Bed level change, shear stress, habitat suitability.
Pitfalls: Picking transport formula outside its regime; ignoring cohesive
sediments.
8) Urban stormwater, CSO control & green infrastructure (LID)
What/Why: Design pipes, storage, and LIDs to reduce flooding and pollution.
Core: EPA-SWMM (runoff + dynamic wave in pipes + LID units).
Advanced: Continuous simulation with climate ensembles, multi-objective
optimization (flooding vs CSO vs cost), real-time control (RTC) with rule-based
or MPC, ML emulators for fast design space search.
Outputs: Node surcharge, CSO volumes, LID treatment performance.
Pitfalls: Calibrating only to total volumes (miss peaks), neglecting backwater.
9) Water distribution networks (pressure, quality, energy)
What/Why: Ensure pressures, minimize leakage/energy, chlorine residuals.
Core: EPANET (steady/extended period) with water age/chlorine decay.
Advanced: Leak detection via pressure residual analytics / GPs, pump
scheduling via MILP/DP, booster chlorination optimization, uncertainty in
demands via ensembles.
Outputs: Pressure maps, critical links, energy use, residual compliance.
Pitfalls: Single-pattern demands; ignoring valve statuses/transient events.
10) Reservoir system operation & allocation (DSS)
What/Why: Balance municipal/irrigation/hydropower/env flows.
Core: Mass balance, rule curves; multi-reservoir dynamic programming or
stochastic DP.
Advanced: Multi-objective NSGA-II fronts (reliability vs. power vs. spills),
reinforcement learning for adaptive rules, scenario trees for climate risk.
Outputs: Optimal releases, shortages, shadow prices.
Pitfalls: Using single historical trace; no hedging during droughts.
11) Climate-change impact assessment
What/Why: Stress-test systems for future rainfall/temperature/ET changes.
Core: Downscaled/bias-corrected GCM/RCM forcings → hydrologic/hydraulic
models.
Advanced: Delta change vs quantile mapping, non-stationary IDF curves,
storylines & robust decision making (RDM) rather than point forecasts.
Outputs: Future floods/droughts, reliability shifts, adaptation portfolios.
Pitfalls: Blindly chaining models without propagating uncertainty.
12) Ecohydrology & environmental flows
What/Why: Link hydraulics/hydrology to habitat and biology.
Core: Habitat suitability indices, PHABSIM/2D ecohydraulic models.
Advanced: Life-stage specific requirements, agent-based or IBM for fish,
connectivity metrics; coupling with reservoir operations to meet E-flows.
Outputs: Habitat-weighted usable area, flow recommendations.
Pitfalls: Static habitat curves used in morphologically dynamic rivers.
13) Real-time forecasting systems (end-to-end)
What/Why: Operational warnings & dashboards.
Stack: Ingest (radar/IMD/ERA5), model chain, data assimilation, ensemble
post-processing (EMOS), CI/CD for models, web maps.
Advanced: Cloud/GPU pipelines, hybrid physics-ML nowcasting (e.g.,
ConvLSTM rainfall → hydraulic emulator).
Outputs: Probabilistic warnings with lead-time skill.
Pitfalls: No reliability calibration; poor fail-safe behavior.
14) Uncertainty, sensitivity & identifiability (cross-cutting)
Why: Decisions need ranges, not single numbers.
Methods: GLUE, Sobol/Morris, MCMC (DREAM), EnKF for state uncertainty.
Deliverables: Prediction intervals, parameter posteriors, sensitivity ranking,
Value-of-Information for new gauges.
Pitfalls: Over-confident single run; ignoring structural error.
15) Digital twins & surrogate modeling
What/Why: High-fidelity + real-time control/what-ifs.
How: Couple calibrated simulators with emulators (Gaussian Processes,
XGBoost, physics-informed NN), continuously synced with sensors/SCADA; use
MPC for pumps/gates.
Gains: Millisecond responses for optimization and operations.
Pitfalls: Emulator drift; training outside design space.
 Advanced modeling applications to
Environmental Engineering problems.

1) Air-quality & Exposure Modeling (urban to regional)

Purpose: Predict pollutant concentrations (PM₂.₅, NOx, O₃, VOCs), source contributions,
and human exposure.

How it’s modeled

 Eulerian chemistry–transport: ∂ C /∂ t+∇ ⋅(u C)=∇ ⋅(K ∇ C)+ R(C , T , rad )+ S−L
(advection, turbulent diffusion, chemical reactions, sources/sinks)
 Lagrangian/Plume for near-source (AERMOD, CALPUFF).
 CFD (LES/RANS) for street canyons, indoor/outdoor coupling.

Data & tools: Emission inventories, meteorology (WRF), land use, chemistry mechanisms
(CB6, SAPRC), monitors/low-cost sensors. Tools: CMAQ/WRF-Chem, CAMx, AERMOD,
OpenFOAM.

Advanced: Chemical data assimilation (EnKF/4D-Var), source apportionment

(PSAT/OSAT), hybrid physics-ML emulators, personal exposure models with mobility
traces.

Outputs/metrics: Concentration maps, exceedance hours, population exposure, source

contributions; RMSE, bias, FAC2.

Pitfalls: Bad emission factors, ignoring secondary aerosols, coarse urban canopy
representation.

2) Water-quality Modeling in Rivers & Lakes

Purpose: Dissolved oxygen, BOD, nutrients (N/P), algae/cyanotoxins, temperature.

How it’s modeled

 Advection–dispersion–reaction:
∂ C /∂ t+ v ⋅ ∇ C=∇ ⋅(D ∇ C)+ R(C , T , light )+S with Monod/first-order kinetics,
light/temperature limitation.
 1D river chains (QUAL2K), 2D/long-and-deep (CE-QUAL-W2), 3D
(WASP/Delft3D-WAQ).
Data & tools: Discharge, temperatures, loads from WWTP/CSO, point & non-point sources,
bathymetry, meteorology.

Advanced: Sediment diagenesis/benthic fluxes, scenario analysis for TMDLs, Bayesian

calibration with multiple constituents, algal functional groups, machine-learned surrogate to
speed scenario sweeps.

Outputs/metrics: DO sag curves, chlorophyll-a peaks, compliance probabilities; NSE, KGE,

percent bias.

Pitfalls: Calibrating only on DO, ignoring residence-time changes and heat budget.

3) Groundwater Flow, Transport & Remediation

Purpose: Drawdown impacts, saltwater intrusion, contaminant plumes (nitrate, PFAS,

metals).

How it’s modeled

 Flow (MODFLOW): S s ∂ h /∂t=∇ ⋅(K ∇ h)+W

 Transport (MT3DMS/RT3D/SEAWAT): advection–dispersion–reaction; density
coupling for salinity.

Data & tools: Heads, pumping, lithology, hydraulic tests; chemistry, sorption isotherms.

Advanced: Dual-domain mass transfer, geostatistical K fields, MCMC uncertainty,

optimization (NSGA-II) for well placement/pumping & barriers, travel-time/capture zones
(MODPATH).

Outputs/metrics: Drawdown cones, capture fractions, cleanup time, mass discharge.

Pitfalls: Using uniform K, Peclet violations (numerical dispersion), calibrating only to heads
(add fluxes).

4) Urban Stormwater & CSO/LID Design (SWMM)

Purpose: Flood reduction and water-quality control in combined/separate sewers.

How it’s modeled:

Rain–runoff (CN/Green–Ampt), kinematic/dynamic wave routing in pipes, storage nodes,
LID units (bioretention, permeable pavement).

Advanced: Continuous multi-year simulations with climate ensembles, real-time control

(RTC) rules or model predictive control, multi-objective optimization (flooding vs CSO vs
cost), ML emulators for fast design.

Outputs: Node surcharge, overflow volumes, event mean concentrations, LID performance.

Pitfalls: Ignoring backwater, calibrating only on volumes not peaks, wrong impervious
connectivity.
5) Solid Waste & Landfill Process Modeling

Purpose: Gas (CH₄/CO₂) generation and collection, leachate quantity/quality, liner integrity.

How it’s modeled:

First-order decay of degradable organics; unsaturated flow in waste (Richards), heat
generation/transfer, gas transport through layered media.

Advanced: Coupled thermo-hydro-biochemical (THBC) models, CFD for gas extraction

field, optimization of wellfield vacuum, LCA coupling for net GHG.

Outputs: Gas yield curves, capture efficiency, leachate loads.

Pitfalls: Poor waste composition data, anisotropy of waste mass ignored.

6) Wastewater Treatment Plant (WWTP) Modeling & Control

Purpose: Predict effluent BOD, NH₄⁺, NO₃⁻, PO₄³⁻, sludge; optimize aeration and
chemicals.

How it’s modeled:

ASM1/2d/3 (activated sludge), secondary clarifier settling (1D layer models), sidestream
processes (Anammox). For digesters: ADM1.

Advanced: Plant-wide models, soft sensors, MPC for DO/air valves, energy–effluent Pareto
optimization, ML anomaly detection.

Outputs: Time series of effluent, energy use, sludge age, carbon/nutrient removal.

Pitfalls: Using generic kinetics/stoichiometry; not calibrating clarifier parameters.

7) Contaminated Sites & Risk Assessment

Purpose: Human/ecological risk from soil/groundwater/vapor intrusion; select remediation.

How it’s modeled:

Multimedia fate & transport (soil–GW–indoor air), exposure pathways, dose–response;
probabilistic (Monte-Carlo) risk.

Advanced: Bayesian updating with new sampling, source identification inversion, value-of-
information to plan new wells.

Outputs: Excess lifetime cancer risk, HQ, cleanup volumes & costs.

Pitfalls: Deterministic single-value screening; ignoring vapor intrusion stack effects.

8) Sediment Transport & Morphodynamics (Rivers/Coasts)

Purpose: Channel stability, dredging needs, scour at piers, beach nourishment.

How it’s modeled:
Saint-Venant + Exner bed evolution; bedload/suspension formulas (Meyer-Peter–Müller,
Engelund–Hansen); morphodynamic acceleration.

Advanced: 2D/3D models (Delft3D/TELEMAC), cohesive sediment & floc dynamics, bank
erosion, habitat coupling.

Outputs: Bed level change maps, shear stress, scour depth.

Pitfalls: Using wrong transport law for grain size/regime; neglecting cohesive behavior.

9) Life-Cycle Assessment (LCA) & Techno-Economic Analysis (TEA)

Purpose: Cradle-to-grave environmental burdens and costs of technologies/policies.

How it’s modeled:

Process flow diagrams → inventory → impact assessment (TRACI/ReCiPe) → uncertainty.

Advanced: Consequential LCA, hybrid EIO-LCA, dynamic LCA (time-varying grids),

TEA-LCA integrated optimization.

Outputs: GWP, eutrophication, acidification, cost per functional unit, sensitivity.

Pitfalls: System boundary creep; database mismatch with local context.

10) Water Distribution Systems (EPANET + Optimization)

Purpose: Pressures, water age/chlorine, energy; leak mitigation.

How it’s modeled:

Mass & head balance in pipes and tanks; chlorine decay kinetics.

Advanced: Demand uncertainty ensembles, leak localization via pressure residuals/GPs,

MILP pump scheduling, booster chlorination optimization.

Outputs: Pressure maps, head losses, residuals, energy KPIs.

Pitfalls: Wrong demand patterns, ignoring valve statuses and transients.

11) Climate-Change Impact & Robust Adaptation

Purpose: Stress-test systems to future rainfall/temperature/ET changes.

How it’s modeled:

Downscale/bias-correct GCM/RCM → hydrologic/hydraulic/water-quality models →
performance under scenarios.

Advanced: Non-stationary IDF curves, storylines (conditioned extremes), Robust Decision

Making (RDM), Info-Gap, multi-model ensembles with reliability calibration.
Outputs: Change in reliability/flood risk, adaptation portfolios with robustness scores.

Pitfalls: Treating one GCM as “truth”; not propagating uncertainty through chains.

12) Ecohydrology & Environmental Flows

Purpose: Link hydraulics to habitat/biota; set e-flows.

How it’s modeled:

Habitat suitability curves (depth/velocity/substrate), PHABSIM, 2D ecohydraulic models;
population dynamics.

Advanced: Agent-based models, life-stage requirements, coupling with reservoir operation

optimization.

Outputs: Weighted usable area, flow recommendations, trade-offs with

hydropower/irrigation.

Pitfalls: Static habitat curves in morphing channels; no validation with biotic data.

13) Remote Sensing + Data Assimilation

Purpose: Fill observation gaps and update model states.

How it’s modeled:

Sensor fusion (rainfall, soil moisture, water surface, temperature, chlorophyll) with
EnKF/Particle Filter into hydro/WQ/groundwater models.

Advanced: Physics-informed neural nets (PINNs), spatiotemporal kriging, bias-aware

assimilation.

Outputs: State estimates with uncertainty bands; improved forecasts.

Pitfalls: Uncorrected sensor bias; filter divergence with wrong error covariances.

14) Decision Support, Multi-Objective Optimization & MCDA

Purpose: Pick portfolios (e.g., LIDs/reservoir rules/WWTP upgrades) under constraints.

How it’s modeled:

Couple simulators with NSGA-II/III, PSO, GA, robust optimization; rank with
AHP/TOPSIS/OWA.

Advanced: Surrogate (Gaussian Process/XGBoost) to accelerate search, stochastic

constraints, equity metrics, real options analysis.

Outputs: Pareto fronts (cost vs risk vs ecology), ranked alternatives, robustness charts.

Pitfalls: Surrogates trained outside design space; ignoring stakeholder weights.

15) Digital Twins for Environmental Assets

Purpose: Real-time “living” models for WWTPs, rivers, stormwater networks, or aquifers.

How it’s modeled:

Calibrated physics + continuous telemetry + online state estimation + control logic
(MPC/RTC). ML emulators provide millisecond responses.

Advanced: Drift monitoring, active learning (acquire new data where uncertainty is high),
anomaly detection.

Outputs: Live dashboards, predictive maintenance alerts, optimized actions.

Pitfalls: Data latency/quality, untracked config changes, cyber-resilience.