36 Highway Engineering

Example 2.7 Contd

1275 1450
1 2 3

325 1975 645

4 5 6
1450 1025

Figure 2.2 Link volumes arising from ‘all-or-nothing’ traffic assignment procedure.

Table 2.14 Minimum time/cost paths between zones in transport network

Destination zone

Origin zone 1 2 3 4 5 6

1 1-2 1-2-3 1-4 1-2-5 1-2-3-6

2 2-1 2-3 2-5-4 2-5 2-5-6
3 3-2-1 3-2 3-2-5-4 3-2-5 3-6
4 4-1 4-5-2 4-5-2-3 4-5 4-5-6
5 5-2-1 5-2 5-2-3 5-4 5-6
6 6-3-2-1 6-5-2 6-3 6-5-4 6-5

Table 2.15 Trip

Destination zone interchanges between
the six zones
Origin zone 1 2 3 4 5 6

1 250 100 125 150 75

2 300 275 200 400 150
3 150 325 100 100 240
4 200 150 50 350 125
5 100 300 125 250 200
6 150 150 180 225 175

2.8 A full example of the four-stage transportation modelling process

2.8.1 Trip production

Assume a study area is divided into seven zones (A, B, C, D, E, F, G) as indi-

cated in Fig. 2.3. Transport planners wish to estimate the volume of car traffic
for each of the links within the network for ten years into the future (termed
the design year).
Using land use data compiled from the baseline year on the trips attracted to
and generated by each zone, together with information on the three main trip
generation factors for each of the seven zones:
 Forecasting Future Traffic Flows 37

Population (trip productions)

Retail floor area (trip attractions)

Employment levels (trip attractions)
linear regression analysis yields the following zone-based equations for the two
relevant dependent variables (zonal trip productions and zonal trip attractions)
as follows:
P = (3 ¥ population) - 500 (2.17)
A = (3 ¥ number employed) + (75 ¥ office floor space, m2) + 400 (2.18)
Table 2.16 gives zonal trip generation factors for the design year, together with
the trip productions and attractions estimated from these factors using Equa-
tions 2.17 and 2.18.

Table 2.16 Trip productions and attractions for the design year (10 years after baseline year)

Zone Population Office floor Numbers Trip Trip

area (m2) employed productions attractions

A 7 500 50 775 22 000 6 475

B 4 000 400 3 500 11 500 40 900
C 6 000 75 700 17 500 8 125
D 5 000 250 4 000 14 500 31 150
E 9 000 100 1 000 26 500 10 900
F 6 000 50 3 000 17 500 13 150
G 4 000 100 800 11 500 10 300
Total 41 500 1025 13 775 121 000 121 000

For example, in the case of zone A:

Trips produced = 3 ¥ 7500 - 500 = 22 000
Trips attracted = (3 ¥ 775) + (75 ¥ 50) + 400 = 6475

2.8.2 Trip distribution

In order to compile the trip distribution matrix, the impedance term relating to
the resistance to travel between each pair of zones must be established. In this
case, the travel time is taken as a measure of the impedance and the zone-to-
zone times are given in Table 2.17.
Using a gravity model with the deterrence function in the following form
between zone i and zone j:
Fij = t -2ij
where tij is the time taken to travel between zone i and zone j
The interzonal trips are estimated using Equation 2.3. For example, taking the
trips from zone A to all other zones, it can be seen from Table 2.16 that 6475
38 Highway Engineering

Table 2.17 Interzonal

Destination zone travel times
Origin zone A B C D E F G

A 10 15 15 20 25 32
B 10 7 5 10 15 22
C 15 7 8 14 16 26
D 15 5 8 6 10 18
E 20 10 14 6 16 12
F 25 15 16 10 16 12
G 32 22 26 18 12 12

Table 2.18 Gravity model computations for Zone A

PF
i ij Aj PF
i ij

Zone Aj Pi Tij Fij Pi ¥ Fij  (PF ) i ij  (PF )i ij

j j

A 6475 22 000
B to A 11 500 10 0.010 115.0 0.317 2053.0
C to A 17 500 15 0.004 77.78 0.214 1388.5
D to A 14 500 15 0.004 64.44 0.178 1150.5
E to A 26 500 20 0.003 66.25 0.183 1182.7
F to A 17 500 25 0.002 28.00 0.077 499.80
G to A 11 500 32 0.001 11.23 0.031 200.50
S = 362.7 S =1 S = 6475

trips were attracted to zone A. Equation 2.3 is used to estimate what propor-
tion of this total amount sets out from each of the other six zones, based on
the relative number of trips produced by each of the six zones and the time taken
to travel from each to zone A. These computations are given in Table 2.18.
When an identical set of calculations are done for the other six zones using
the gravity model, the initial trip matrix shown in Table 2.19 is obtained.
It can be seen from Table 2.19 that, while each individual column sums to
give the correct number of trips attracted for each of the seven zones, each indi-

Table 2.19 Initial output from gravity model

Destination zone

Origin zone A B C D E F G Total

A 5 905 1019 1 713 740 958 525 10 861

B 2053 2446 8 060 1 547 1391 581 16 078
C 1388 9 587 4 791 1 201 1861 633 19 461
D 1150 15 569 2361 5 418 3947 1 094 29 540
E 1183 7 113 1409 12 898 2818 4 498 29 919
F 500 2 088 712 3 066 920 2 970 10 256
G 200 638 177 622 1 074 2174 4 886
Total 6475 40 900 8125 31 150 10 900 13 150 10 300 121 000
 Forecasting Future Traffic Flows 39

vidual row does not sum to give the correct number of trips produced by each.
(It should be noted that the overall number of productions and attractions are
equal at the correct value of 121 000.)
In order to produce a final matrix where both rows and columns sum to their
correct values, a remedial procedure must be undertaken, termed the row-
column factor technique. It is a two-step process.
First, each row sum is corrected by a factor that gives the zone in question
its correct sum total (Table 2.20).
Second, because the column sums no longer give their correct summation,
these are now multiplied by a factor which returns them to their correct indi-
vidual totals (Table 2.21).
This repetitive process is continued until a final matrix is obtained where the
production and attraction value for each zone is very close to the correct row
and column totals (Table 2.22).

Table 2.20 Row correction of initial gravity model trip matrix

Destination zone
Origin Correct Row
zone A B C D E F G Total total factor

A 5 905 1019 1 713 740 958 525 10 861 22 000 2.026

B 2053 2446 8 060 1 547 1 391 581 16 078 11 500 0.715
C 1388 9 587 4 791 1 201 1 861 633 19 461 17 500 0.899
D 1150 15 569 2361 5 418 3 947 1 094 29 540 14 500 0.491
E 1183 7 113 1409 12 898 2 818 4 498 29 919 26 500 0.886
F 500 2 088 712 3 066 920 2 970 10 256 17 500 1.706
G 200 638 177 622 1 074 2 174 4 886 11 500 2.354
Total 6475 40 900 8125 31 150 10 900 13 150 10 300 121 000 121 000

Table 2.21 Column correction of gravity model trip matrix

Destination zone
Origin
zone A B C D E F G Total

A 0 11 962 2064 3 470 1 499 1 941 1 064 22 000

B 1468 0 1750 5 765 1 107 995 415 11 500
C 1249 8 621 0 4 308 1 080 1 673 569 17 500
D 565 7 642 1159 0 2 660 1 938 537 14 500
E 1048 6 301 1248 11 424 0 2 496 3 984 26 500
F 853 3 562 1216 5232 1 569 0 5 068 17 500
G 472 1 501 417 1 464 2 529 5 117 0 11 500
Total 5654 39 589 7854 31 663 10 443 14 160 11 636
Correct
total 6475 40 900 8125 31 150 10 900 13 150 10 300 121 000
Column
factor 1.145 1.033 1.035 0.984 1.044 0.929 0.885
40 Highway Engineering

Table 2.22 Final corrected trip matrix

Destination zone

Origin zone A B C D E F G Total

A 12 286 2112 3 352 1 551 1 780 918 22 000

B 1670 0 1800 5 599 1 152 918 361 11 500
C 1407 8 818 0 4 144 1 114 1 528 489 17 500
D 632 7 759 1172 0 2 722 1 757 458 14 500
E 1222 6 673 1317 11 380 0 2 360 3 548 26 500
F 998 3 784 1286 5 226 1 680 0 4 526 17 500
G 547 1 579 437 1 448 2 681 4 807 0 11 500
Total 6475 40 900 8125 31 150 10 900 13 150 10 300 121 000

(Note, if Equation 2.2 is used within the trip distribution process, the rows
sum correctly whereas the columns do not. In this situation the row-column
factor method is again used but the two-stage process is reversed as a correc-
tion is first applied to the column totals and then to the new row totals.)

2.8.3 Modal split

Two modes of travel are available to all trip makers within the interchange
matrix: bus and private car. In order to determine the proportion of trips under-
taken by car, the utility of each mode must be estimated. The utility functions
for the two modes are:
UCAR = 2.5 - 0.6C - 0.01T (2.19)
UBUS = 0.0 - 0.6C - 0.01T (2.20)
where
C = cost (£)
T = travel time (minutes)
For all travellers between each pair of zones:

The trip by car costs £2.00 more than by bus

The journey takes 10 minutes longer by bus than by car.
Since the model parameters for the cost and time variables are the same in
Equations 2.19 and 2.20, the relative utilities of the two modes can be easily
calculated:
U(BUS -CAR ) = (0.0 - 2.5) - 0.6(c - (c + 2)) - 0.01((t + 10) - t )
= -2.5 + 1.2 - 0.1
= -1.4
 Forecasting Future Traffic Flows 41

U(CAR-BUS ) = (2.5 - 0.0) - 0.6((c + 2) - c) - 0.01(t - (t + 10))

= 2.5 - 1.2 + 0.1
= 1.4
where
£c = cost of travel by bus
£(c + 2) = cost of travel by car
t = travel time by car (in minutes)
(t + 10) = travel time by bus (in minutes)
We can now calculate the probability of the journey being made by car using
Equation 2.16:

PBUS = 1 ∏ (1 + e(UCAR -UBUS ) )

= 1 ∏ (1 + e(1.4 ) )
= 0.198
PCAR = 1 ∏ (1 + e(UBUS -UCAR ) )
= 1 ∏ (1 + e( -1.4 ) )
= 0.802
So just over 80% of all trips made will be by car. If we assume that each car
has, on average, 1.2 occupants, multiplying each cell within Table 2.22 by 0.802
and dividing by 1.2 will deliver a final matrix of car trips between the seven
zones as shown in Table 2.23.

Table 2.23 Interzonal

Destination zone trips by car
Origin zone A B C D E F G

A 0 8213 1412 2241 1037 1190 614

B 1117 0 1203 3743 770 613 241
C 940 5895 0 2771 744 1022 327
D 422 5187 784 0 1820 1174 306
E 817 4461 880 7607 0 1578 2372
F 667 2529 860 3494 1123 0 3025
G 366 1056 292 968 1792 3213 0

2.8.4 Trip assignment

The final stage involves assigning all the car trips in the matrix within Table 2.23
to the various links within the highway network shown in Fig. 2.3. Taking the
information on the interzonal travel times in Table 2.17 and using the ‘all-or-
nothing’ method of traffic assignment, the zone pairs contributing to the flow
along each link can be established (Table 2.24). The addition of the flows from
each pair along a given link allows its 2-way flow to be estimated. These are
shown in Fig. 2.4.
42 Highway Engineering

C F

A D G

B E

Figure 2.3 Zones and links in study area within worked example.

Figure 2.4 Interzonal link flows for private vehicles (cars).

Table 2.24 2-way vehicular flows along each link

Network link Zone pairs contributing to flow along link Total link flow

A to B (A,B)(B,A) (A,D)(D,A) (A,E)(E,A) (A,F)(F,A) (A,G)(G,A) 16 683

A to C (A,C)(C,A) 2 352
B to C (B,C)(C,B) 7 098
B to D (A,D)(D,A) (A,F)(F,A) (B,D)(D,B) (B,F)(F,B) 16 592
B to E (A,E)(E,A) (A,G)(G,A) (B,E)(E,B) (B,G)(G,B) 9 362
C to D (C,D)(D,C) (C,E)(E,C) (C,G)(G,C) 5 798
C to F (C,F)(F,C) 1 882
D to E (C,E)(E,C) (C,G)(G,C) (D,E)(E,D) (D,G)(G,D) 12 944
D to F (A,F)(F,A) (B,F)(F,B) (D,F)(F,D) 9 667
E to F (E,F)(F,E) 2 701
E to G (A,G)(G,A) (B,G)(G,B) (C,G)(G,C) (D,G)(G,D) (E,G)(G,E) 8 334
F to G (F,G)(G,F) 6 238

2.9 Concluding comments

The process of traffic forecasting lies at the very basis of highway engineering.
Modelling transport demand is normally undertaken using a four-stage sequen-
tial process starting with trip generation and distribution, followed by modal
 Forecasting Future Traffic Flows 43

split and concluding with traffic assignment. Predicting flows along the links
within a highway network provides vital information for the economic and envi-
ronmental assessments required as part of the project appraisal process and
allows the scale of each individual project within the network to be determined.
Once the demand analysis and appraisal process have been completed, the
detailed junction and link design can then be undertaken.
It should be remembered, however, that the modelling process is a simplifi-
cation of reality. Predictions arising from it are broad estimates rather than
precise forecasts. The error range within which the model results are likely to
fall should accompany any data supplied to the transport planners.

2.10 References

Furness, K.P. (1965) Time function iteration. Traffic Engineering Control, 7, 458–460.
McFadden, D. (1981) Economic models and probabilistic choice. In Structural Analysis
of Discrete Data with Econometric Applications (eds Manski & McFadden). MIT Press,
Cambridge, MA, USA.
Wardrop, J.G. (1952) Some theoretical aspects of road traffic research. Proceedings of the
Institution of Civil Engineers, 1(36), 325–362.