CHAPTER 7

TRIP DISTRIBUTION

7.1 Basis of Trip Distribution

After having obtained an estimate of the trips generated from and attracted to the various
zones, it is necessary to determine the direction of travel. The number and types of trips
originating and terminating in different zones of the study area have been estimated in the first
phase of the modeling process (Trip generation). It is known now for what purpose these trips
are to be made. The purpose of the trip distribution analysis phase is to develop a procedure that
synthesizes the trip linkages between traffic zones for both transit captive and choice trip makers.
Figure 7.1 shows that knowledge of the properties of the alternative horizon - year transport
networks is required for the trip distribution analysis phase.
If Oi is the number of trips having their origin in zone i and Dj is the number of trips
having terminated in zone j, trip distribution stage determines the number of trips Tij which
would originate from zone i and terminate in zone j. A trip matrix is drawn for the study area
with the sum of rows indicating the total number of trips (Oi) originated in zone i and the sum of
columns the total number of destinations (Dj) or attractions to zone j.
If an area is divided into n number of zones, possible zone to zone movements will be n
x n and the trip matrix will be as shown in figure 7.1. The purpose of the procedure is to
complete each cell entry Tij so that the sum of rows and columns are as nearly equal as possible
to the total future trips in the study area.

Figure 7.1
1: Propertiees required of
o synthesizzed trip disttribution maatrices

7.2 P-A Matrix

M
to O-D
O Matrix
7.2.1 Bacckground
Once
O
estimate of the trip generated, i.e.
i producedd from and aattracted to thhe various zzones,
are madee the next steep is to deteermine the direction
d
of ttravel of theese trips. Thee number off trips
produced
d in any zon
nes of the stu
udy area hass to be appo rtioned to thhe zones to which thesee trips
are attraccted. Say, theere are pi trip
ps produced
d from zone i by trip makkers categoryy q and aj is the
number of
o trip ends attracted to zone j, then
n the numberr of trips beetween zone i and zone j (i.e.
tqi-j) wou
uld be estimated using trip
t distributtion techniqu
que. This cann be represeented in a m
matrix
form as given
g
below::
1

1 t11 t12 t13 .. t1n

2 t21 t22 t23 ..
t2n
n

pi
p1
p2

tn1
tn2 tn3 ..
pn
n
tnn

aj a1 a2

a3 an

The horizontal axis of the trip matrix shown represents the zones of attractions (i.e.
destinations) and the vertical axis represents the zones of production (i.e. origins) numbered from
1 to n. The total of any individual row, i, represents the total number of trips produced in zone i
(pi) and the total of any individual column j represents the number of trips attracted to zone j (aj).
An Origin-Destination (O-D) trip table is a two dimensional matrix of elements whose
cells values represents the travel demand between each given origin (row) and destination
(column) zone. An O-D trip table can be obtained by conventional surveys such as license plate
surveys, home interviews, roadside surveys etc. Such surveys are time consuming, expensive and
labor intensive. In addition, many of these approaches involve sampling errors. These
conventional approaches also suffer from other drawbacks, such as an inability to reflect changes
in influencing factors. For instance, if the land use characteristics change, so will the trip table.
Hence the previously measured trip table may quickly become outdated, and one needs to repeat
these surveys in order to obtain new trip tables, which is cost prohibitive.
Nevertheless, transportation organizations may require the trip tables for planning
purposes. Due to constraints in budget, time and labor, researchers began exploring alternative
means of producing these trip tables. Since the early 1970s several techniques have been
explored to obtain the trip table without the need for expensive surveys, as will be described
further in the literature review.

7.2.2 Needs for O-D Matrix

Traffic congestion is one of the major problems faced by both developed and developing
countries. Several potential solutions have been investigated to solve it, and a great deal of effort
is put into activities such as, transportation planning to transportation management in order to
attempt to alleviate the problem. Some solutions include: demand restriction policies (such as
increasing taxes or increasing other transportation expenses, ramp metering), increasing
transportation supply, improved transportation management strategies (such as better signal
coordination, real time traffic diversion) and recently development of ITS (Intelligent
Transportation Systems). Most transportation problems are complicated. However, solutions to
them are costly when put into practice, prohibiting trial and error approaches to finding the best

solution, as a careful investigation is needed to understand the effect of relevant demand and
supply before any strategy is implemented.
Demand information in transportation planning is described by trip tables. There are two
types of matrices; Production-Attraction (P-A) matrix and Origin-Destination (O-D) matrix. The
main difference between the two is that the cell values in a P-A trip table are non directional,
whereas the cell values of an O-D trip table have directional meaning, indicating the number of
trips going from an origin to a destination. Furthermore, a P-A trip table satisfies flow
conservation, i.e. the summation of productions must be equal to summation of attractions,
whereas an O-D trip tables does not exhibit such a property.
O-D trip tables are generally used for traffic assignment purposes. During the past few
years many transportation engineers have begun to focus on Intelligent Transportation Systems
(ITS), which include Advanced Traffic Management Systems (ATMS), Advanced Traveler
Information Systems (ATIS) and Automated Vehicle Control Systems (AVCS). It is believed
that accurate and fast O-D trip table generation techniques are needed, in the planning, operation
and maintenance of such ITS.

7.3 Gravity Model of Trip Distribution:The Gravity model gets its name from the fact that it is conceptually based on Newtons
law of gravitation, which states that the force of attraction between two bodies is directly
proportional to the product of the masses of the two bodies and inversely proportional to the
square of the distance between them. With analogy from physical science, the law of gravitation
was applied to estimate the flow between two trip potential sources. Thus the form is

I G
ij

PP
d
i

ij

The first standard model for general application was formulated by BPR as

T ODF

Where,

Oi = Origin totals from zone i

Dj = Destination totals to zone j

Fij = Travel time of friction factor which is a function of deterrence i.e f(dij)
The row sums (origin constraint) have been satisfied within the model form by specifying the
model in the form

T O
ij

DF
j

D F
j

T O
j

ij

ij

ij

The model can be formed form the argument that probability of trip interchange is influenced by

a) The trip production of origin zone i

b) The trip attraction of destination zone j
c) The separation between i & j
The trip production is reflection of zone size and number of surveyed trip origins Oi . It is the
row total of Tij matrix.

The trip attraction is also the reflection of size of the zone in terms of the number of surveyed
trip destinations Dj. It is the column totals of the Tij matrix.
Influence of the separation of the origin and destination as explained above i.e (Fij).

Thus Tij

Oi

Dj

Fij

This is the gravity model in the most elementary form. This, as it stands would not suffice for
several reasons
The product Oi Dj in the equation would mean that if Oi & Dj are doubled there would
be four times as many trips. But often trip matrix of say 50 x 50 size for a city would be
aggregated to a 25 x25 matrix for analytical convenience, but the trips for the city as a whole
would remain same. Also, as it stands there would be no means to determine Fij .
While the form in above equation will be retained, a number of constraints will be
introduced to overcome the problems. These are

1.

T O
ij

2.

T D
i

ij

3. TLFD (survey) = TLFD (model)

4. TLFD = Trip Length Frequency Distribution

We achieve these constraints in following ways:

Satisfaction of Constraint 1:
The model equation can be changed to

O D F
i

ij

D F

ij

ij

The first constraint is achieved automatically as below

T O
ij

D F
j

D F

It should be noted that

D *F
j

A
i

ij

ij

ij

is a constant for each row and can be written as

1
D F
j

ij

Satisfaction of Constraint 2:
This is achieved by successive iterations whereby the value of Dj adjusted till constraint 2
is satisfied. The Dj value is given in table 7.1. Each i j cell of the tables contains the following:

Fij
Dj

Fij

Oi. Dj .Fij

Dj. Fij

Table 7.1: Dj Values

Zones

Dj (1)
ITERATION(1)

60

37

33

19

59

Dj (2)
ITERATION(2)

50.44

38.35

42.79

20.89

59.81

Dj (3)
ITERATION(3)

50.15

38.45

43.26

21.59

59.07

Example zone-2

7.3.1 Calibration of gravity model

The model is

T
ij

ODF
i

ij

D F
j

ij

and the denominator assures the satisfaction of the constraint

T O
ij

The other two Constraints to be satisfied in the process of Calibration are

T D
i

ij

& TLFD (survey/ observed) = TLFD (Modelled)

An iterative procedure is required to satisfy these two constraints.

Phase - 1:

ODF

ij

[Iteration-1]

D F

ij

ij

Phase - 2: Add over the columns of the

T S
ij

ij

matrix

j(1)

Sj (1) = is the column sum after iteration 1.

Phase - 3: If Sj (1) are not some as survey Dj then modify Dj for next iteration.

D
S
j

j( 2 )

j(1)

j(1)

Phase - 4:

T
ij

OD F
i

j( 2 )

T S
ij

Phase - 6:

j( 3)

ij

[Iteration-2]

D F
j

Phase - 5:

j( 2 )

ij

j( 2 )

D
D
S

j( 2 )

j( 2 )

Phase - 7:

T
ij

OD F
i

j( 3)

ij

D F
j( 3)

ij

In general terms

T
ij

OD F
i

ij

D F
j

Where,

j( k )

j( k )

ij

[Iteration - 3]

j( k )

D
D
S
j

j( k 1)

j( k 1)

7.3.2 Case Studies of Gravity Model:Length frequency distribution for the base year. A gravity model of the type Figure -1
shows a simple five-zone city along with the trip productions, trip attractions and the trip defined
in equation below may be calibrated for this simple city.

T O
ij

DF
j

ij

D F
j

ij

Figure 7.5 shows the relationship between the travel time factor and the travel time
assumed for the first iteration of the model. It should be noted for this simple city that there are
only two origin zones, numbers 3 and 5. Table 7.2 shows the calculations required by the first
iteration of the model. The entries in lines 1, 2, 6, and 7 of table 7.2 are obtained directly from
the information provided in figure 7.6 and 7.7. The Sum of Dj Fij products in lines 3 and 8 are the
denominators of the gravity model for zones 3 and 5 respectively.

Figure 7.5

Figure 7.6 Travel ch

haracteristiccs of a simplle city
This
T denomin
nators is usu
ually referred
d to as the aaccessibility of the particcular origin since
it convey
ys the accesss which thaat zone has to opportunnities in thee destinationn zones. Forr this
simply ciity, househo
olds located in
i zone 3 haave greater aaccessibility to employm
ment opportunities
than do households
h
lo
ocated in zone 5.
The
T entries in
n lines 4&9 are measurees of the atttraction that a destinatioon zone has to an
origin zo
one. It should
d be noted th
hat the entriees in these linnes sum to 11.00 since prroductions frrom a
zone of origin must all find destinations. The
T six Tijss calculated in table 7.22 along witth the
associateed Dijs may be used to construct
c
the simulated trrip-length frrequency disstribution andd this
is shown on fig 7.7. Also
A shown on fig 7.7 iss the observeed frequencyy distributionn.

Figure 7.7
7: Characteriistics of firstt Iteration off gravity model for sampple city
A Compariso
on of these frequency distributions
d
indicates thhat the gravvity model iis not
distributiing sufficien
nt trips to th
he shorter travel time. E
Estimate of the travel tiime factors to be
used for the second iteration & figure 7.7. Travel timee factors haave been inccreased for ttravel
m
thiss will have th
he effect in tthe gravity m
model of disstributing a llarger
times of less than 3 minutes,
number of
o trips to deestinations with
w shorter trravel times.

The
T calculatiions presentted in table--1(below) coould now be repeated w
with the neew Fij
magnitu
udes and a seecond trip len
ngth frequen
ncy distributtion.
Calculaations requireed for first itteration

TABLE 7.2
7
Calculatio
ons

Destinatiion Zone
1

Attraction
ns(Oi)

450

250

3300

Origin 3

1.Travel tiime Dij

0.36
0

0.45

0..18

2.Travel tiime factor Fij

162
1

112

554

328

3.Dj Fij

0.50
0

0.34

00.16

1.00

150
1

102

448

300

7. Travel time
t
factor Fij

0.36
0

0.18

0..25

8. Dj Fij

162

45

775

282

9. Oi Fij / D
j Fij

1.57

0.16

0..27

1.00

399

112

1889

700

4.Oi Fij / D
j Fij
5.

T O
ij

DF
D F
j

ij

ij

Origin 5
6. Travel time
t
Dij

10.

T O
ij

DF
D F
j

i
ij

ij

7.4 Grow
wth Factor Methods
M
of Trip Distriibution:Growth
G
factors methods are
a generally
y simple andd have been used in earllier studies. Their
basic assumptions are projected to
t the design
n year in the future by ussing certain expansion fa
factor.
Various growth
g
facto
or methods which
w
have been
b
developped from tim
me to time arre given beloow

i)

Uniform Growth
G
Facttor Method

ii)

Average Growth Factor Method

iii)

Detroit Method

iv)

Frater Method

v)

Furness Method

7.4.1 Uniform growth Factor Method

This is the oldest method and a Single growth factor, F is calculated for the entire zone.
The future trips between zone i & j are then calculated by using equation

T T F
h

ij

ij

Where h and b stand for horizon year and base year respectively and F is the growth
factor. Let us take a numerical example to illustrate the Procedure
Example: The trip interchanges among three zones of a study area are shown in the matrix
below
1

50

25

25

50

100

75

25

100

150

25

75

150

The growth factors estimated for different zones over the 20 year design period are as given
below. Estimate the trip matrix for the design year.
Zone:

G.F :

Solution: Future trip matrix will simply be forecasted by multiplying each row by its growth
factor as shown below:

150

75

75

200

400

300

25

100

150

50

150

300

This method of trip distribution suffers from numbers of brawbacks. The assumption of
uniform growth rate for the entire zone is not rational. The method underestimates the
movements where present day trips are intensive.

7.4.2 Average Factor Method

In this method an average of two zonal growth factors is applied to the base year
trips to arrive at a forecast for the design year.

T t
h

ij

i j

F F
2
i

Where Fi & Fj are growth factors for zone i&j respectively.

Taking the Previous example future trips between zone 1 and 2 may be calculated as follws.

F F
2
3 4
50
175
2

T t
h

12

12

Similarly,

T 150
h

3y

1 2
225and so on
2

the resulting matrix is given in table 7.3

Table 7.3
1

Oi

Oi

New G.F

required
1

175

50

62.5

287.5

300

1.043

175

250

225*

650

900

1.384

50

250

225

525

275

0.524

62.5

225*

225

512.5

500

0.975

As may be seen average of two growth factors has underestimated some trips and over
estimated. Others zone a new growth factor is to be calculated by taking the ratio of Oi
estimated and process is to be repeated till the completed values match with the required
values.

7.4.3 Fratar Growth Factor Method:

Several naive trend or simple growth factor models have been developed for use in
special situations. Among these the Fratar model is often used to estimate external trips. That
71

is trips that are either produced and/or are attracted outside the boundaries of the region under
study from outlying areas whose character is not explicitly analyzed. The Fratar growthfactor method uses the following expression to synthesize horizon year trip- interchange
magnitudes:

h T b F F Li L j
T ij
i
j
ij
2
T bij
j
Where, Li
T bij F j
j
Lj

T b
ij
i

T b
ij F i
i

h = The number of vehicle trips between zones i& j in the horizon year.
T ij
Tb
ij = The number of vehicle trips between zones i & j observed in the base year.

F i , F j = the growth factor for zones i& j which reflect the growth in trip

productions and

trip attractions expected between the base and horizon years where

Li , L j = the locational factors

N.B: The locational factors defined the reciprocals of the average attracting forces of all of
the surrounding zones.
Application of Fratar model
Mathematical expressions are:

h T b F F Li L j
T ij
i
j
ij
2
T bij
j
Where, Li
T bij F j
j
Lj

T b
ij
i

T b
ij F i
i

72

Taking an example
1

25

50

25

25

150

75

50

150

200

25

75

200

The growth factors estimated for different zones over the 20 year design period are as
given below.

Zone: 1 2

G.F: 3 4

Using the above equation we get

25 50 25
0.44
(25 4) (50 2) (25 1)

25 150 75
0.55
(25 3) (150 2) (75 1)

L1

12

L2
2

(0.44 0.55) 2 0.50

25 3 4 0.50 150

Similarly all other values are calculated. The matrix obtained after 1st iteration will be as
follows:
1

150

130

31

311

300

0.96

587

192

929

1000

1.07

162

879

800

0.91

385

300

0.78

150

130

587

31

192

162

reqd

Modified G.F

For second iteration

150 130 31
1.03
(150 1.07) (130 0.91) (31 0.78)

150 587 192

1.12
(150 0.96) (587 0.91) (192 0.78)

73

130 587 162

0.99
(130 0.96) (587 1.07) (162 0.78)

31 192 162
1.00
(31 0.96) (192 1.07) (162 0.91)

After second iteration we make a matrix and one more iteration may be carried out for
further refinement in the results.
Disadvantage

i.

It is not sensitive to changes in the properties of transport network.

ii.

Changes in the behavior of trip makers in forming linkages between various types of
land-based cannot be reflected in this trip distribution model.

iii.

For those cities in which significant changes in urban structure are not expected in
this model.

iv.

It breaks down mathematically when a new zone is built after the base year since all
base year interchange volumes involving such a zone would be equal to zero.

7.4.4 Disadvantage of Growth Factor Method:

The major disadvantage of growth factor method are- that in each method present
day trip distribution matrix has to be obtained first and any error in that will get magnified in
future trip matrix. They do not take into account any changes in the accessibility during the
forecast period.
Conclusion:

Despite the above shortcomings, the growth factor methods are relatively simpler to
use and understand. They can be used for studies of small areas for updating stable and
uniform data.

74