-----------o0o----------

TRANSPORTATION PLANNING

Trip Distribution
Presented by
Huy Huu Nguyen. PhD

Ho Chi Minh City, April 2020

1
 Contents
 Introduction

 Planning Concepts and Four-Step Process Overview

 Trip Generation

 Trip Distribution

 Mode Choice

 Traffic Assignment

 Travel Demand Model Applications

2
 Trip Distribution

 Trip distribution is the second step in the travel forecasting process. This step
determines where the generated trips are going to and coming from.
 Trip distribution establishes the desire lines between those zones and connects the
trips from any one zone to all other zones.
 The final product of the trip distribution step is what is known as an origin –
destination matrix, or O/D matrix.
 The O/D matrix is a large table that lists all origins as rows and all destinations as
columns. The table entries then refer to the number of trips estimated between
each zonal pair.
 It is important to emphasize that trip distribution makes no assignment of how traffic
is expected to get from one zone to the other (i.e., which roads do people choose to
get from A to B), but rather estimates the total expected flow between A and B,
independent of travel path.

3
 Trip Distribution

 In the distribution step, the trips originating from and destined to a zone are
generally referred to as productions and attractions, or Ps and As.
 In large travel demand models, the result of the trip distribution step is also referred
to as a trip table, which contains the O/D matrix, as opposed to a path table, which
is the result of traffic assignment.
 The trip distribution step is applicable to small planning studies, or TIAs, as well as
long-term planning studies using travel demand models (TDMs). For a TIA, trip
distribution estimates the trips from the site under study to or from each exit from
the study area. For a TDM, trip distributions estimates the trips between each zone
and every other zone.
 Trip distribution data (O/D patterns) can be estimated from field data collection,
synthesized (combined) from existing turning movement counts, or estimated
through analytical approaches. The most common of those approaches is called
the gravity model.

4
 Trip Distribution
4.1 Gravity Model
 The gravity model is an analytical methodology to estimate the productions and
attractions between each zonal pair in the trip distribution step. It is most applicable for
larger studies and networks, and can be quite data intensive to complete.
 The underlying concept for the gravity model mirrors principles in physics, where two
masses attract one another, and where the magnitude of that attraction is proportional to
the size of the mass.

 In other words, larger zones (or more precisely, zones with more activity or higher
development density) attract more trips than smaller zones (less activity and a lower
development density). That physics-based gravity model is further calibrated by the
relative travel time between zones and socioeconomic characteristics.

5
 Trip Distribution
4.1 Gravity Model
 Specifically, the gravity model estimates the number of trips between each origin
zone, i, and each destination zone, j, as a function of the following characteristics:

• The number of trips produced in each zone, Pi;

• The number of trips attracted to each zone, Aj;
• A friction factor for each ij zonal pair that is a function of the travel time
between those zones, Fij;
• A socioeconomic factor for each ij zonal pair that can be used to calibrate the
desire of travelers to travel between those zones, Kij.

Fij

Kij

6
 Trip Distribution
4.1 Gravity Model
 The general form of the gravity model is shown in the following equation, followed
by an example.

 The gravity model has to be applied in an iterative fashion for both the total attractions
and productions for each zone to match with the entered data. In the first iteration, the
productions for each zone will be balanced, but the attractions will likely differ from the
entered trip generation data. In repeated iterations, the results of one iteration are fed
as input to the next iteration, until a desirable match is obtained.

7
 Trip Distribution
4.1 Gravity Model
EXAMPLE 2.8
A new office park is expected to generate 1500 homebound trips in the PM peak
hour. Analysts expect the trips to terminate in four residential zones.

Using a gravity model, estimate the number of trips to go from the office park to
zone 3 during the PM peak hour?

8
 Trip Distribution
4.1 Gravity Model
Solution (EX 2.8)
This problem is best solved by setting up a table to calculate the trips from the office park to
each zone. In this example, Pi = 1500 for all rows in the table. The table shows attractions
for each destination zone, estimates friction factors as the inverse of travel time, and gives
the socio-economic factor. Those three terms are used to estimate the A  F  K product for
each destination zone, which are then used to get the estimate of the number of trips for
each zonal pair.

 The total number of trips estimated from the office park to zone 3 is 146 trips.

9
 Trip Distribution
4.2 Use of Data and Calibration
 Trip distribution data (origin/destination or O/D patterns) can be estimated from field
data collection, although the data collection effort can be extensive.
 Traditional O/D survey methods include license plate matching surveys and market
analysis surveys.
 License-plate matching surveys can be performed manually (noting the last three
digits of a plate is often sufficient) or can be automated using automated license
plate recognition (ALPR) technology.
 Modern Bluetooth reader technology can also be used to estimate O/D patterns by
recording Bluetooth MAC signatures of mobile devices (cell phones, tablets, etc.) in
one location and matching them with readings at other locations.

10
 Trip Distribution
4.2 Use of Data and Calibration

 Bluetooth studies and license plate studies are generally sample based, meaning
that only a subset of all traffic is observed to get O/D percentages, which are then
extrapolated to the full O/D matrix by scaling the matrix with actual traffic counts.
 O/D studies are very time consuming and labor intensive, as the data collection
units (ALPR, cameras, Bluetooth) need to be deployed simultaneously at all major
entry and exit points to the network, as well as at whatever internal zones are
desired.
 In practice, these O/D surveys are therefore often limited to key origin – destination
pairs or used as a tool to calibrate an estimated O/D table.

11
 Trip Distribution
4.2 Use of Data and Calibration
 For small networks, the O/D matrix can also be synthesized from existing intersection
turning movement counts (% left, % through, % right) by assuming proportionality.
 For a corridor study, this process may start at a downstream intersection. If we assume
an east-to-west corridor, the turning percentages at the eastbound approach at the
eastern-most intersection can be assumed to be proportionally fed by all trip origins at
the next upstream intersection (in this case eastbound through, northbound right, and
southbound left). Figure 2.8 illustrates this concept for a one-way street in the eastbound
direction.

Figure 2.8 O/D

synthesis for a
corridor study

12
 Trip Distribution
4.2 Use of Data and Calibration
 In part (a), the figure shows turning movement counts. The gray highlighted cells for
southbound right (120 veh/h), eastbound through (450 veh/h), and northbound right (60
veh/h) form the combined traffic stream of 630 veh/h traveling toward intersection 2. The
contribution of the three streams to that 630 veh/h are 19%, 71%, and 10%, respectively.
At intersection 2, the eastbound approach shows left-turn, through, and right-turn
volumes of 85, 500, and 45 veh/h, respectively. Because the network is balanced, this
also adds up to 630 veh/h.

Figure 2.8 O/D

synthesis for a
corridor study

13
 Trip Distribution
4.2 Use of Data and Calibration
 To estimate the origin – destination pattern, the (destination) turning movement counts
at intersection 2 are multiplied by the three origin percentages from intersection 1. For
example, the 85 left-turning vehicles at intersection 2 are divided into 19% from
intersection 1 north approach (16 veh/h), 71% from intersection 1 west approach (61
veh/h), and 10% from intersection 1 south approach (8 veh/h).

Figure 2.8
O/D
synthesis for
a corridor
study

14
 Trip Distribution
4.2 Use of Data and Calibration
 The same process is applied for the other movements and the results are shown in part
(b) of the figure.
 Finally, the origins at intersection 1 and the destination 2 are summed internally, to make
sure that all the calculations yield consistent volumes with the original turning movement
counts.

Figure 2.8
O/D
synthesis for
a corridor
study

For a longer corridor, this process starts at the most downstream end (e.g.,
intersection 5 for a five-intersection corridor) and then works its way upstream
toward intersection 1.
15
 Trip Distribution
4.3 Sample Network Application of Trip Distribution
 Figure 2.9 shows a sample network with eight internal TAZs (part of the study area)
and four external TAZs (not explicitly being studied, but contribute to trip in and out of
study area).
 The figure further shows a roadway network with classification into principal arterials/
freeway, principal arterial/surface street, and minor arterials.
 Collector streets and local roads are omitted from the network in a planning context
in this example, but can be included if they carry significant amounts of traffic.
 As a rule of thumb, the total traffic on nonmodeled links should be less than 10% of
the total traffic in the network.
 The figure shows estimated trips between zone 1 and all other zones as weighted
desire lines, with the weight of each line being proportional to the magnitude of
forecasted traffic between each zonal pair.

16
 Trip Distribution
4.3 Sample Network Application of Trip Distribution

Figure 2.9 Sample network: Trip distribution step

17
 Trip Distribution
4.3 Sample Network Application of Trip Distribution
 Sample results from the trip distribution step are shown in Table 2.3 for zone 1. These
results would be repeated for all other zones in a planning study
 For each zone, the table gives the total trips attracted by the remaining 13 zones from the
trip generation step. It then shows the travel times between zones (TTij, in minutes),
which are converted into friction factors between zones by taking the inverse.
 The table then shows the estimated socioeconomic factors, assigning larger weights for
pairs of zones with more interzonal trips.
 The table then shows the sum of attractions, friction factors, and socio-economic factors
for each destination zone, which are summed at the bottom of the table.
 The second to last column divided each “A  F  K” product by the total sum. In the last
column, this product is multiplied by the total trips produced in zone 1 (750 trips, see
Table 2.2) to estimate the total trips between zone 1 and the remaining 13 zones.

18
 Trip Distribution
4.3 Sample Network Application of Trip Distribution
Table 2.3 Sample result of trip distribution step

19
 Trip Distribution
4.3 Sample Network Application of Trip Distribution
 The trip distribution step is repeated for the remaining 13 zones, to arrive at a similar
table of trips between zones 2 through 14 and the remaining zones. The results of the
overall trip distribution step are often displayed in the form of an origin – destination
matrix (O/D matrix), which is also referred to as a trip table.
 The resulting O/D matrix for the example described after one iteration of the gravity
model is shown in Table 2.4.

 At this point it is again emphasized that multiple iterations of the gravity model are
needed to arrive at a balanced O/D matrix or trip table. This is very evident in Table 2.4,
in which the row totals from the trip distribution (column “row sum”) match the input data
from the trip generation step.
 In other words, each zone produces the correct number of trips as was estimated in that
prior step. However, the column totals for each destination zone do not necessarily
match, with most destination zone totals (“col sum”) being significantly less than the field
data.
20
 Trip Distribution
4.3 Sample Network Application of Trip Distribution
Table 2.4 Origin – Destination matrix after one iteration of gravity model

 Consequently, additional iterations of trip distribution are needed, where the results of
iteration 1 are used as inputs for iteration 2, and so forth. Sometimes, many iterations are
needed for both the row and column totals to converge, which is especially the case for
large networks.

21
 Trip Distribution

Figure 2.8
O/D
synthesis for
a corridor
study

22
 Trip Distribution

23
 Trip Distribution

Visualization of Trip Distribution Results

24
THANK YOU
Nguyen Huu Huy. PhD
Department of Transport Planning

Contact:
Phone: 0939.098.090
E-mail: huycongtrinh@hcmutrans.edu.vn
Website: www.ut.edu.vn

25