INDIAN

 INSTITUTE  OF  SCIENCE  

STOCHASTIC HYDROLOGY
Lecture -23
Course Instructor : Prof. P. P. MUJUMDAR
Department of Civil Engg., IISc.
Summary  of  the  previous  lecture  

• Markov chains
– Transition probabilities
– Transition probability matrix (TPM)
– Steady state Markov chains

2  
 Markov Chains
• stochastic process with the property that value of
process Xt at time t depends on its value at time t-1
and not on the sequence of other values
P [ X t X t −1 , X t −2 ,..... X 0 ] = P [ X t X t −1 ]
P ⎡⎣ X t = a j X t −1 = ai ⎤⎦ = Pijt
(n) (0) n
p = p ×P
• At steady state,
p = p× P
3  
 Example – 1
Consider the TPM for a 2-state first order homogeneous
Markov chain as
⎡0.7 0.3⎤
TPM = ⎢ ⎥
⎣ 0.4 0.6 ⎦
State 1 is a non-rainy day and state 2 is a rainy day
Obtain the
1. probability that day 1 is a non-rainy day given that day 0 is
a rainy day
2. probability that day 2 is a rainy day given that day 0 is a
non-rainy day
3. probability that day 100 is a rainy day given that day 0 is a
non-rainy day
4  
 Example – 1 (contd.)
1. probability that day 1 is a non-rainy day given that day
0 is a rainy day
No rain rain
No rain ⎡0.7 0.3⎤
TPM =
rain ⎢ 0.4 0.6⎥⎦
⎣
The probability is 0.4

2. probability that day 2 is a rainy day given that day 0 is

a non-rainy day
p ( 2 ) = p (1) × P
p(1), in this case is [0.7 0.3] because it is given that
day 0 is a non-rainy day.
5  
 Example – 1 (contd.)
( 2) ⎡0.7 0.3⎤
p = [0.7 0.3] ⎢ ⎥
⎣ 0.4 0.6 ⎦
= [0.61 0.39]
The probability is 0.39

3. probability that day 100 is a rainy day given that day 0

is a non-rainy day

(n) (0)
p = p × Pn

6  
 Example – 1 (contd.)
P2 = P × P
⎡0.7 0.3⎤ ⎡0.7 0.3⎤ ⎡ 0.61 0.39 ⎤
= ⎢ ⎥ ⎢ ⎥ = ⎢ ⎥
⎣ 0.4 0.6 ⎦ ⎣ 0.4 0.6 ⎦ ⎣ 0.52 0.48 ⎦
4 2 2 ⎡0.5749 0.4251⎤
P = P × P = ⎢ ⎥
⎣ 0.5668 0.4332 ⎦
8 4 ⎡ 0.5715
4 0.4285 ⎤
P = P × P = ⎢
⎣0.5714 0.4286 ⎥⎦
16 8 8 ⎡0.5714 0.4286 ⎤
P = P × P = ⎢
⎣0.5714 0.4286 ⎥⎦

7  
 Example – 1 (contd.)
Steady state probability
p = [0.5714 0.4286]

Verification: at steady state,

p = p× P
⎡0.7 0.3⎤
= [0.5714 0.4286] ⎢ ⎥
⎣ 0.4 0.6 ⎦
= [0.5714 0.4286]

8  
 Markov Chains
Data generation:
• Data generation from a Markov chain needs
knowledge of the initial state and the transitional
probability matrix (TPM).
• To determine the state at time 2, get a random
number Ru between 0 and 1. n −1 n
• If this random number Ru is between ∑ pij and ∑ pij
j =1 j =1

for n = 1, 2, ……m, the next state is taken as state

n

9  
 Example – 2
Consider a Markov chain model for annual streamflow at
a location. Assume the TPM as
⎡0.7 0.3 0 ⎤
TPM = ⎢⎢ 0.1 0.6 0.3⎥⎥
⎢⎣ 0.5 0.4 0.1⎥⎦

State 1 – Deficit flow

State 2 – Intermediate Flow
State 3 – Excess flow (Flow above a threshold)

10  
 Example – 2 (Contd.)
Consider that the flow is in state 1 at t = 0; generate a
sequence of 10 possible flow states corresponding to t =
1, 2, ……..10

The cumulative transition probability matrix (P*) is

⎡0.7 1 1⎤
P* = ⎢⎢ 0.1 0.7 1⎥⎥
⎢⎣ 0.5 0.9 1⎥⎦

11  
 Example – 2 (Contd.)
Time State at t Ru State at t+1
1 1 0.896 2
2 2 0.919 3
3 3 0.682 2
4 2 0.922 3
5 3 0.735 2
6 2 0.435 2
7 2 0.006 1
8 1 0.154 1
9 1 0.549 1
10 1 0.612 1
12  
 Example – 3
Consider that last two days weather conditions influence
the today s weather condition. i.e., whether or not today
is a dry day or a wet day depends on the previous two
days weather condition.

Consider,
•If the past two days have been wet days, then it will be a
wet day tomorrow with a probability 0.8
•If today is a wet day but yesterday was a dry day, then it
will be a wet day tomorrow with probability 0.5
•If yesterday was a wet day but today is a dry day, then
tomorrow will be a wet day with probability 0.3
13  
 Example – 3 (Contd.)
• If the past two days have been dry days, then
tomorrow will be a wet day with probability 0.1

If we let the state at time t depend only whether or not

it is a wet day at time t , then the preceding model is
not a single step Markov chain

However, it is possible to transform this model into a

Markov chain by saying that the state at any time is
determined by the weather conditions during both that
day and the previous day.

14  
 Example – 3 (Contd.)
i.e., State 1 – If both today and yesterday are wet days
State 2 – If today is a wet day but yesterday is a dry day
State 3 – If yesterday was a wet day but today is a dry
day
State 4 – If both today and yesterday are dry days

t-2 t-1 t
State 1 W W
State 2 D W
State 3 W D
State 4 D D

15  
 Example – 3 (Contd.)
• If the past two days have been wet days, then it will be
a wet day tomorrow with a probability 0.8

t-2 t-1 t-1 t p

State 1 W W W W 0.8
State 2 D W D W 0
State 3 W D W D 0.2
State 4 D D D D 0

[0.8 0 0.2 0]

16  
 Example – 3 (Contd.)

The preceding would then represent a 4 state Markov

chain with transition probability matrix as

⎡0.8 0 0.2 0 ⎤
⎢0.5 0 0.5 0 ⎥
TPM = ⎢ ⎥
⎢ 0 0.3 0 0.7 ⎥
⎢ ⎥
⎣ 0 0.1 0 0.9 ⎦

17  
 Example – 3 (Contd.)
Consider that Monday and Tuesday are wet days, what is
the probability that Thursday is a wet day?

Given that Monday and Tuesday are wet days, the

probability vector of Wednesday is

p(W ) = [0.8 0 0.2 0]

(T ) (W )
p =p ×P

18  
 Example – 3 (Contd.)
⎡0.8 0 0.2 0 ⎤
⎢0.5 0 0.5 0 ⎥
p (T ) = [0.8 0 0.2 0] ⎢ ⎥
⎢ 0 0.3 0 0.7 ⎥
⎢ ⎥
⎣ 0 0.1 0 0.9 ⎦

= [0.64 0.06 0.16 0.14]

Thursday: Wet Wet Dry Dry

Therefore the probability of Thursday being a wet

day = 0.64+0.06 = 0.7
19  
 Markov Chains
Difficulties in using Markov chains in hydrology
• Determining the number of states to use.
• Determining the intervals of the variable under
study to associate with each state.
• Assigning a number to the magnitude of an event
once the state is determined.
• Estimating the large number of parameters
involved in even a moderate size Markov chain
model.
• Handling situations where some transitions are
dependent on several previous time periods while
others are dependent on only one prior time period.
Ref: Statistical methods in Hydrology by C.T.Haan, Iowa state university press 20  
FREQUENCY ANALYSIS
 Frequency Analysis
• Hydrologic systems are influenced by severe
events occurring.
– e.g., severe storms, floods, droughts.
• The magnitude of an extreme event is inversely
proportional to its frequency of occurrence (i.e.,
severe events occur less frequently).
• Frequency analysis is a procedure for estimating
the frequency (or the probability) of occurrence of
past and (or) future events

22  
 Frequency Analysis
• Objective of frequency analysis of hydrologic data
is to relate the magnitude of extreme events to their
frequency of occurrence using probability
distributions.
• Hydrologic data to be analyzed is assumed to be
independent and identically distributed and the
hydrologic system is assumed to be stochastic,
space-independent and time independent
• The data should be properly selected so that the
assumptions of independent and identically
distributed are satisfied.

23  
 Frequency Analysis
• The assumption of identical distribution or the
homogeneity is achieved by selecting the
observations from same population
• The assumption of independence is achieved by
selecting the annual maximum of the variable being
analyzed as the successive observations from year
to year will be independent.

24  
 Frequency Analysis
• The results of flood frequency analysis can be used
in many engineering purposes
– E.g., flood flow frequency analysis can be used
in the design of dams, bridges, culverts, flood
controlling devices
– Urban flooding

25