Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

116 Transportation Planning—Principles, Practices and Policies

Average trip length by the simulated matrix = 2.79 km


Average trip length by the observed matrix = 2.79 km
Hence, by taking the value of a = 2.3, we get the average trip lengths of the simulated
matrix and the observed matrix equal. Therefore, the value of calibrated parameter a for gravity
model is 2.3.

6.3.2 Opportunity Model


Opportunity models are based on the statistical theory of probability as the theoretical foundation.
The concept has been pioneered by Schneider and developed by the subsequent studies.
The two well-known models are given below:
1. Intervening opportunities models
2. Competing opportunities model
Opportunity model can be represented by the following general formula:
Ti–j = OiP(Dj)
where, Ti–j is the predicted number of trips from zone i to j, Oi is the total number of trips
originating in zone i, P(Dj) is the calculated probability of a trip terminating in zone j, and Dj
is the total of trip destinations attracted to zone j.
Intervening opportunities models
The original form of intervening opportunity model was proposed by Stouffer[3]. Intervening
opportunity model assumes that the trip interchange between an origin and a destination zone
is equal to the total trips emanating from the origin zone multiplied by the probability that each
trip will find an acceptable terminal at the destination. It is further considered that the probability
that a destination will be acceptable is determined by two zonal characteristics—first, the size of
the destination and second, the order in which it is encountered as trips proceed from the origin.
The probability equation P(Dj) may then be presented as the difference between the probability
that the trip origins at i will find a suitable terminal in one of the destinations, ordered by the
closeness to i, up to and including j, and the probability that they will find a suitable terminal
in the destinations up to but excluding j. Equation (6.11) represents mathematically this concept:
Ti–j = Oi (e–LB – e–LA) (6.11)
where, Ti–j is the predicted number of trips from zone i to j, Qi is the total number of trips
originating in zone i, L is the probability density (probability per destination) of destination
acceptability at the point of consideration, A is the number of destinations between i and j
(including j) when arranged in order of closeness and B is the number of destinations between
i and j (excluding j) when arranged in order of closeness.
Also, A = B + Dj (6.12)
The model is developed on the assumption that the probability of choice of a particular
destination (from a given origin of a particular trip purpose) is proportional to the opportunities
at the destination and inversely proportional to all such opportunities that are closer to the origin.
The inverse proportionality to closer opportunities can be expressed as the proportionality to
the probability that none of the closer destinations (opportunities) are chosen[5]. Schneider
Trip Distribution Models 117

first used it extensively in a transportation study by the Chicago Area Transportation Study
(CATS) and has since been, used by several other transportation studies for possible destinations
of a specific trip. The model is conceptualised on the basis of two assumptions or hypotheses
about human behaviour. These are
1. Total travel time from a point is minimised when subjected to the condition that
every destination point has a stated probability of being accepted, if it is considered.
2. The probability of a destination being accepted, if it is considered, is a constant and
independent of the order in which destinations are considered.
Consider an individual at home who decides that he needs certain type of sweets. He may
go to any sweet shop. He considers a sweet shop in terms of increasing travel time from his
home. The probability that the individual will accept any one sweet shop is M. This assumption
is that there are no differences in the quality, availability, price, etc. of the sweets needed so that
each sweet shop is perceived as being equally acceptable as any other sweet shop. Assuming
that the individual tries to minimise his travel time, the alternative destinations can be considered
in travel time order from home. In general, the probability of going to anyone destination is
the probability that the destination is an acceptable one, multiplied by the probability that the
individual has not already stopped at a preceding destination. The probability of accepting a
destination is denoted M.
Considering the first sweet shop in the example, the probability of acceptance is M and
the probability that the individual has not already stopped at a preceding sweet shop is l (since
there is no any preceding sweet shop). Thus, the probability PM that he stops for the first time
is given by Eq. (6.13):
Pf = M × 1 (6.13)
Now, considering the second sweet shop, there is a probability (1 – M) that the individual did
not stop at the first store and a probability M that the second store is acceptable. Hence, the
probability of stopping at the second store is given by Eq. (6.14):
Ps = (1 – M)M (6.14)
Like the gravity model, the intervening opportunity model also experiences drawbacks with
respect to its calibration process.
Derivation of the model: Schneider proposed a modification of the Stouffer hypothesis [3].
The model is derived as follows using the concept of Schneider:
jµ = µth destination in the rank list from a particular origin i
Tijµ µ opportunities,
that is, the probability that the individual will proceed beyond the µth destination
M
Djµ = number of opportunities at the µth destination
if µ = 1, we can formulate the following:
Tij1
In probability terms, this is the same as
Tij1
118 Transportation Planning—Principles, Practices and Policies

Then, Eq. (6.15) can be expressed mathematically as


Tij1 = 1 – MDj1 (6.15)
When µ = 2,
Tij2 = probability that an individual will proceed beyond the second destination
This can also be stated as:
Tij2 = probability that the individual is not satisfied with the first two destinations
These can be expressed as a combined probability.
Tij2 = probability that the individual is not satisfied with the first destination (Qij1)
× probability that the individual is not satisfied with the second destination
Now, we know that
Probability that the individual is not satisfied with the second destination
= 1 – Probability that the individual is satisfied by the second destination
= 1 – MDj2
Further, Eq. (6.16) is obtained with the same argument expressed earlier.
Tij2 = Tij1 (1 – MDj2) (6.16)
when µ = 3, and following the same procedure, we obtain
Tij3 = Tij2 (1 – MDj3) (6.17)
and so on in general,
Tijµ = Tijµ–1 (1 – MDjµ) (6.18)
Equation (6.18) can be written as
Tij Tij 1
MD j (6.19)
Tij 1

If Vjµ is the number of opportunities passed up to and including zone jµ, then
Djµ = Vjµ – Vjµ–1 (6.20)
Therefore, Eq. (6.19) becomes
Tij Tij 1
M (V j Vj 1) (6.21)
Tij 1

Equation (6.21) can be written, with the assumption of continuous variation, as


dT
MdV
T
which integrates to give
In T = –MV + constant
T = K exp(–MV)
So that in discrete notation, is can be expressed as follows:
Tijµ = Ki exp(–MVjµ) (6.22)
Trip Distribution Models 119

where Ki is a constant.
Equation (6.22) represents that the proportion of trips proceeding beyond jµ varies
exponentially with the cumulative number of opportunities.
The number of trips from zone i to the µth destination away is obtained as follows:
Iijµ = Total number of trips originating at I(Oi) × probability that a trip will remain
at the µth destination (6.23)
We know that the number of trips from zone i is likely to be terminated at zone j which will
be equal to the probability of trip from zone i terminating at zone j, multiplied by the number
of trips originating in i.
Probability that a trip will remain at the µth destination = Probability that a trip will
proceed beyond the (µ – 1)th destination – Probability that a trip will proceed beyond the µth
destination.
Thus, Eq. (6.23) can be expressed mathematically as follows:
Iijµ = Oi(Tijµ–1 – Tijµ)
Substituting Eq. (6.23) in Eq. (6.22) we obtain Eq. (6.24).
Iijµ = KiQi[exp(–MVjµ–1) – exp(–MVjµ)] (6.24)
This is the expression of the intervening opportunities model. In certain critical states, the
intervening opportunities model approximates to the gravity model.
EXAMPLE 6.1 The number of trips produced in or attracted to the three zones 1, 2 and 3
are presented in Table 6.57.

TABLE 6.57 Number of Trips Produced/Attracted


Zone 1 2 3 Total
Trips produced 16 35 30 81
Trips attracted 35 30 16 81

The order of closeness of the zones is shown in the following matrix (Table 6.58).
TABLE 6.58 Order of Closeness of Zones
D 1 2 3
O
1 1 2 3
2 2 1 3
3 2 3 4

The zonal factors are given in Table 6.59.


TABLE 6.59 Zonal Factors
Zone Zonal factors
1 0.06
2 0.03
3 0.06
120 Transportation Planning—Principles, Practices and Policies

Distribute the trips between the zones.


Solution Using Eq. (6.25) given below,
Ti–j = Oi(e–MB – e–MA) (6.25)
We get
T1–1 = 16[e–0.06 × 0 – e–0.06 × 35] = 14
T1–2 = 16[e–0.06 × 35 – e–0.06(35 + 30)] = 2
T1–3 = 16[e–0.06 × (30 + 35) – e–0.06(35 + 30 + 16)] = 1
T2–1 = 35[e–0.03 × 30 – e–0.03(35 + 30)] = 13
T2–2 = 16[e–0.03 × 0 – e–0.03 × 30] = 21
T2–3 = 35[e–0.03 × (35 + 30) – e–0.03(35 + 30 + 16)] = 2
T3–1 = 30[e–0.06 × 16 – e–0.06(16 + 35)] = 1
T3–2 = 30[e–0.06 × (35 + 16) – e–0.06(16 + 35 + 30)] = 1
T3–3 = 30[e–0.06 × 0 – e–0.06 × 16] = 18
Accordingly, O-D matrix is plotted (Table 6.60) and the number of trips distributed is checked.

TABLE 6.60 O-D Matrix


D
1 2 3 Total
O
1 14 2 1 17
2 9.02 21 1.89 33
3 1 1.71 18.51 20
Destination total 24.02 24.17 21.4 70

It is thus seen that 73 trips out of 81 have been distributed by this stage and further
iterations are needed. The destination total can be adjusted by the formula given below:
[D j D( m 1) ]
D j(m)
C j(m 1)

where, Dj(m) is the adjusted destination total for iteration m, Dj is the desired destination total,
D(m–1) is the adjusted destination total for preceding iteration (m – 1) and Cj(m–1) is the actual
destination total, iteration (m – 1).
The iteration continues till a reasonable closeness is observed between the obtained total
trips and the calculated total trips.
According to Eq. (6.25), adjusted destination total for zone 1, 2 and 3 can be calculated
as under:
35 35
A1 51
24.02
30 30
A2 = = 37.5
24
16 16
A3 11.96
21.4
Trip Distribution Models 121

Taking these values, further iteration needs to be performed, as discussed. Again, a new
set of destination totals has to be worked out through a number of iterations till the desired
destinations match with the actual destinations.
Competing opportunities
This technique has been used in Penn-Jersey Study which takes into consideration the direct
application of probability theory along with some aspects of gravity model and Fratar technique
of successive approximation. In this model, the adjusted probability of a trip ending in a zone
is the product of two independent probabilities[5]. The structure of the model is as under:
Ti–j = Ti(Gi) Pj
where Ti–j is the number of one way trips from zone i to zone j, Ti(Gi) is the total number of
trips originating at zone i and Pj is the adjusted probability of stopping at zone j.
The adjusted probability is defined as the product of two independent probabilities—
probability of attraction and probability of satisfaction.

SUMMARY
Most of the trip distribution models pertaining to growth factor and synthetic models have been
discussed in great detail along with their advantages and disadvantages. Numerical examples of
each model have also been demonstrated to appreciate its merits and demerits. Most popular
model used in transportation studies is the gravity model. Derivation of gravity model based
on theory of entropy maximisation has been discussed with examples. Derivation of intervening
opportunity model along with its basic concept and example has also been described for greater
understanding to the reader. Important part is how accurate is the estimate of trips to be
distributed between the origins and destinations when compared to trips actually distributed
in the real-life situation. It is also seen how close is the shape of the model trip length trip
distribution to the trip length distribution observed through survey. A number of statistical tests
may be carried out to validate the model. One of the statistical tests could be chi-square test,
which examines and compares the cell value of each O-D pair between the model trip matrix
and the survey trip matrix.

REVIEW QUESTIONS
1. Explain the various methods for trip distributions with respect to growth factor models.
2.
3. What is entropy maximisation? Determine which of the following four trip distribution

zones 1 to 2, 3, 4 and 5 is likely to be obtained with the use of entropy maximisation


model.
TABLE 6.61 TABLE 6.62
2 3 4 5 2 3 4 5
1 3 4 8 5 1 4 5 6 5

You might also like