CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Module 2
2.1 Transportation Planning Process
The transportation planning process has a lot of similarity to the problem solving process. The following table gives the major differences between the two processes.

Sl No

Problem Solving

1 2

3 4 5

Transportation Planning Problem definition and Objective Problem solving lacks foresight ness relevant to planning condition. They to take advantage of the forthcoming change themselves, so innovations innovations are used It is not Programmed Basis Usually Programmed basis Our concern may be for the dimension and performance of a We may be concerned with about vehicle to be replaced within a shorter location and capacity of Mass period of time from now Transit Breadth of problem: i.e. parking, Study of broader situation i.e. whole congestion city Immediate solution is required .so it is completed within shorter period Implemented Sequentially

2.2 Types of Planning Methodologies

1. Projective planning. 2. Deductive planning. 3. Objective planning. Projective planning: It is a base for planning. It is an open Extrapolation method. Example: Traffic flows, Vehicle ownership, Residential Densities, Population trends, Economic Growth, Socio- economic indices. Deductive planning: Synthesis the future state of the system from laws, equations or models that are one in its behaviour. Example: Analysis of specific projects and operational activities such as bypasses, regional centers, transport terminals, one-way streets can be effectively analyzed deductive planning process.

18

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay Objective planning: Planner sets some goals and with a certain objective and with constraints. It will be difficult to take into account the uncertainties.

1 Statement of Transportation problems and problem domain

2 Identification of objectives and constraints to be met by proposed transportation system 5

3 Modeling of transportation systems and its Domain

4 Collection data on present transportation system and its domain

Calibration and use of models Projection of levels of Regional land use development and other affecting variables Setting of levels of transportation system variables Prediction of levels of sub regional land use development travel and other affected variables

13

Operation and Maintenance of system 11

Evaluation of performance and decision among alternative systems

Generation and resetting of transportation system variables

Specification of details of selected transportation systems

10 Implementation of system changes 12

Total Transportation planning process

19

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Deductive Planning Process

2.3 Travel demand modelling:

There are four steps of travel demand modelling. They are 1. Trip generation 2. Model split 3. Trip Distribution 4. Trip assignment

20

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Steps of Travel Demand Modelling

2.3.1 Trip-Generation Analysis: -Two types of trip-generation analysis are carried

out and these are trip production and trip attraction. Trip Production: -is reserved for trips generated by residential zones where these trips may be trip origins and destinations. Trip Attraction: -is used to describe trips generated by activities at the non-home end of a home-based trip such as employment, retail service, and so on.

21

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay The first activity in travel-demand forecasting is to identify the various trip types important to a particular transport-planning study. The trip types studied in a particular area depend on the types of transport-planning issues to be resolved. The first level of trip classification used normally is a broad grouping into home-based and non-home-based trips. Home-based Trips: - are those trips that have one trip end at a household. Examples journey to work, shop, school etc. Non-home-based trips: -are trips between work and shop and business trips between two places of employment. Trip classification that have been used in the major transport-planning studies for home-based trips are: a. Work trips b. School trips c. Shopping trips d. Personnel business trips, and e. Social-recreational trips

Factors influencing Trip Production

Households may be characterized in many ways, but a large number of trip-production studies have shown that the following variables are the most important characteristics with respect to the major trip trips such as work and shopping trips: 1. The number of workers in a household, and 2. The household income or some proxy of income, such as the number of cars per household.

Factors Influencing Trip Attraction

Depending on the floor areas, the trip attraction can be determined from retail floor area, service and office floor area and manufacturing and wholesaling floor area.

Multiple Regression Analysis

The majority of trip-generation studies performed have used multiple regression analysis to develop the prediction equations for the trips generated by various types of land use.

22

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay Most of these regression equations have been developed using a stepwise regression analysis computer program. Stepwise regression analysis programs allow the analyst to develop and test a large number of potential regression equations using various combinations and transformations of both the dependent and independent variables. The planner may then select the most appropriate prediction equation using certain statistical criteria. In formulating and testing various regression equations, the analyst must have a thorough understanding of the theoretical basis of the regression analysis.

Review of Regression Analysis Concept Some of the fundamental of regression analysis: - The principal assumptions of regression analysis are:
1. The variance of the Y values about the regression line must be the same for all magnitudes of the independent variables. 2. The deviations of the Y values about the regression line must be independent of each other and normally distributed. 3. The X values are measured without error 4. The regression of the dependent variable Y on the independent variable X is linear. Assume that observation of the magnitude of a dependent variable Y have been obtained for N magnitudes of an independent variable X and that on an equation of the form

Ye = a + bX is to be fitted to the data where Ye is an estimated magnitude rather than an

observed value Y . From the least-squares criterion, the magnitude of the parameters a and b may be estimated.

b=

xy x
2

a = Y bX

where
x = X X and y = Y Y

23

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay X , Y =the means of the X and Y observations respectively.

y
Where

2 = yd + y e2

y
y

= total sum of the squares of the deviations of the Y observations about the

mean value
2 d

=the sum of the squares of the deviations of the Y observations from the

regression line.

2 e

=the sum of the squares of the deviations of the estimated Ye magnitude about

the mean value. The ratio of the sum of the squares explained by the regression to the total sum of squares is known as the coefficient of determination and denoted by r2.

r2 =

y y

2 e 2

0 r2 1

if r 2 =1 implies no variation remaining that is unexplained by the independent variable used in the regression. If r 2 = 0 implies the independent variable used would not explain any of the observed variation in the dependent variable.

The square root of the coefficient of determination is termed as the correlation

coefficient.

A second useful measure of the validity of a regression line is the standard error of the estimate, which is estimated from:

se =

( N 2)

2 d

24

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay where

( N 2) is the degree of freedom associated with the sum of squares

2 d

The regression coefficient b is the statistical estimate and is therefore subject to error.

sb =

se sX

where
s X is the standard deviation of the independent variable.

Statements about the confidence that might be placed in an estimated coefficient is given by: t= regression coefficient standard error of the regression coefficient

Partial or Multiple Regression Equation

It has equation of the form: Ye = a + b1 X 1 + b2 X 2 + ... + b p X p where there are p independent variables and the regression coefficients b1 , b2 ,, b p are referred to as partial regression coefficients. The coefficient of multiple determinations, R 2 2 is given by:

(Y Y ) = (Y Y )
e

where R 2 is known as the multiple correlation coefficients.

se = [ N ( P + 1)]

2 e

25

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay s e2 sb = 2 2 [ s Xi N (1 R Xi )] Where s Xi is the standard deviation of the independent variable X i and R Xi is the coefficient of multiple correlations between X i and all other independent variables.
Table 1: Linear Regression

Y
9428 2192 330 153 3948 1188 240 55 2064 280 52 230 420 9654 450 130

X
9482 2010 574 127 3836 953 223 36 2223 272 50 209 410 11023 527 183

y = Y Y
7502.125 266.125 -1595.88 -1772.88 2022.125 -737.875 -1685.88 -1870.88 138.125 -1645.88 -1873.88 -1695.88 -1505.88 7728.125 -1475.88 -1795.88

x=X X

xy
56066193.42 365.921875 2289482.172 3335885.922 3695180.672 778919.2969 3010340.547 3690534.797 29610.54688 2858267.672 3670218.422 3051939.047 2407329.422 69664216.8 2186693.297 3278594.297

x2
55851334 1.890625 2058149 3540513 1827.375 -1055.63 -1785.63 -1972.63 214.375 -1736.63 -1958.63 -1799.63 -1598.63 9014.375 -1481.63 -1825.63

7473.375 1.375 -1434.63 -1881.63 1827.375 -1055.63 -1785.63 -1972.63 214.375 -1736.63 -1958.63 -1799.63 -1598.63 9014.375 -1481.63 -1825.63

Y
1925.875 b

X
2008.625 2.604143 a -3304.87

=160013772.3

=61445839

Category analysis:
Category analysis is a technique for estimating the trip production characteristics of households, which have been sorted into a number of separate categories according to a 26

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

set of properties that characterize the household. Category analysis may also be used to estimate trip attractions.

Zonal trip productions may be estimated as

piq = hi (c)tp (c)

Where,

piq = The number of trips produced by zone i by type q people.

hi(c)= number of households in zone i in category c tp( c)= trip production rate of a household category c.

Zonal trip-attractions may be estimated as

a j = b j (c)ta (c)

Where, aj = number of work trips attracted by zone j. bj(c) = number of employment opportunities in category c. ta(c) = trip attraction rate of employment category c

And the summation is over all employment types if work trip attractions are to be estimated.

2.3.2 Modal Split

The second stage of travel demand forecasting process has bee identified as captive modal split analysis. The second stage of modal split analysis was identified as occurring after the trip distribution analysis phase. Two submarkets for public transportation services have been labelled as captive transit riders and choice transit riders. The aim of captive modal split analysis is to establish relationships that allow the trip ends estimated in the trip generation phase to be partitioned into captive transit riders and choice transit riders. The purpose of choice modal split analysis phase is to estimate the probable split

27

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

of choice transit riders between public transport and car travel given measures of generalized cost of travel by two modes. The ratio of choice trip makers using a public transport system varies from 9 to 1 in small cities with poorly developed public transport systems to as high as 3 to 1 in well developed cities. Major determinants of Public Patronage are 1. Socio economic characteristics of trip makers 2. Relative cost and service properties of the trip by car and that by public transport. Variables used to identify the status at the household level are 1. Household income or car ownership directly 2. The number of persons per household. 3. The age and sex of household members. 4. The purpose of the trip.

The modal split models, which have been used before the trip distribution phase, are usually referred to as trip end modal split models. Modal split that have followed the trip distribution phase are normally termed trip interchange modal spit models. Trip end modal split models are used today in medium and small sized cities. The basic assumption of the trip end type models is that transport patronage is relatively insensitive to the service characteristics to the transport modes. Modal patronages are determined principally by the socio economic characteristics of the trip makers. Most of the trip interchanges modal split models incorporate measures of relative service characteristics of competing modes as well as measures of the socio economic characteristics of the trip makers. The modal split model developed during the southeastern Wisconsin transportation study is an example of trip end type model. The model-split model developed in Toronto is an example of trip interchange modal split model.

Land Use

28

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Trip Generation equation

Modal split model

Trip Distribution models

Trip ends by mode

Origin-Destination Volumes

Trip Distribution model

Modal split model

O-D Volumes by mode

O-D Volumes by mode

Trip End Type Modal Split Model

Trip Interchange Type Modal Split Model

These factors, including time and cost, can be grouped into three broad categories.

Characteristics of the traveler -- the trip maker; Characteristics of the trip; and Characteristics of the transportation system.

Southeastern Wisconsin Model:

This Model consisted of seven estimating surfaces that related the percentage of the trip ends that will use transit services from a particular traffic analysis zone to the following variables: trip type, characteristics of trip maker and characteristics of the transport system. The trips made by public transport services are classified as Home based work trips, home based shopping trips, home based other trips and non-home based trips. The 29

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

socioeconomic characteristics of trip makers were defined on the zonal basis in terms of average number of cars per household in a zone. The characteristics of a transport system relative to given zone were defined by an accessibility index which is given by:

acci = a j f ij
j =1

Where

acci = accessibility index for zone i a j = No of attractions in zone j f ij = Travel time factor for travel from zone I to zone j for a particular mode being
considered. The transport service provided to a particular zone by two modes was characterized by accessibility ratio.

Accessibility Ratio = Highway accessibility index / Transit accessibility index.

Toronto Model:
Trip interchange modal split models allocate trips between public transport and private transport after trip distribution stage. The split between two modes is assumed to be a function of the following transportation variables between each pair of zones, as well as the socioeconomic characteristics of the people who avail themselves of the alternatives

Relative Travel Time (TTR) Relative Travel Cost Economic status of the Trip maker Relative travel service
X1 + X 2 + X 3 + X 4 + X 5 X6 + X7 + X8

TTR =

30

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Where

X1 = Time spent in transit vehicle X2 =Transfer time between transit vehicles X3 = Time spent in waiting for a transit vehicle X4 = Walking time to a transit vehicle X5 = Walking time for transit vehicle X6 = Auto driving time X7 = Parking delay at destination X8 = Walking time from parking place to destination
This represents the door to door travel time to train to that of automobiles

CR =

X9 (X10 + X 11 + 0.5X 12 )

X 13

Where X9 =Transit Fare X10 =Cost of gasoline X11 =Cost of oil charge and lubrication X12 =Parking Cost at destination X13 = Average Car Occupancy The denominator indicates that auto cost must be put on a person per one way trip basis in order to be comparable to the costs for transit.

The variable EC is defined in terms of medium income per worker in the zone of trip production. The variable D is designed arbitrarily as
D= X2 + X3 + X4 + X5 X7 + X8

MODAL SPLIT MODEL WITH A BEHAVIOURAL ANALYSIS

31

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay The focus of these models is on individual behavior rather than on zonally aggregated modal choice behaviour. Central to these methods is the concept of trip disutility or the generalized cost of using different modes of transport.
Generalized cost of travel:

The concept of generalized travel cost is derived from the notion that the trip making has a number of characteristics which are unpleasant to trip makers and that the magnitudes of this unpleasantness depend on the socioeconomic characteristics of the trip maker. The generalized cost or disutility of a trip may be estimated from:
m m z ij = a n x nij + bw u w + c

n=1,.n, w=1,.w. Where,

m = generalized cost of travel between zones I and j by mode m. zij m xnij = the nth characteristic of mode m between zones I and j which gives rise to the cost

of travel by mode m. u w = the wth socioeconomic characteristic of a tripmaker. c = constant. an , bw = coefficients that reflect the relative contribution that system and tripmaker characteristics make to the generalized cost of travel.

For a binary modal choice situation the following generalized cost difference may be calculated from the above equation.
* z ij = a n x nij + bw u w + c

n=1,.n, w=1,.w. where,

* = the difference in generalized costs of travelling between zones i and j. zij

xnij = the difference in the nth system characteristic between the two modes.

32

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay If trip makers are classified into a number of socioeconomic groups, then may be expressed as
* z ij = a n x nij + c

n=1,,n

* if zij is the difference in cost between transit and car travel then c may be regarded as a

mode penalty reflecting the inferior convenience and comfort of transit relative to car.

Wilson and his associates reported the following generalized cost relationship from studies in England.
m m m m z ij = 0.66d ij + 1.32eij + a 3 sij

Where
m zij = The generalized cost of travelling between zones i and j by modes m m = The in vehicle travel time in minutes by mode m between zones i and j. d ij m eij = The excess travel time in minutes by mode m between zones i and j. m sij = The distance in miles by mode m between zones i and j.

a3 = 2.0 for car travel 2.18 for train travel 3.06 for bus travel.

BINARY CHOICE STOCHASTIC MODAL SPLIT MODELS:

This is one of the models which deal with the generalized costs of travel for competing modes. Three types of mathematical concepts have been used to construct stochastic modal choice functions for the individual behaviour: Discriminant analysis Probit analysis Logit analysis.

33

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Discriminant analysis

The basic premise is that the choice of tripmakers in an urban area may be classified into two groups according to mode of transport used. The objective is to find a linear combination of explanatory variables that possesses little overlap. The best discriminate function is the one that minimizes the number of mis classifications of trip makers to the observed transport modes. Quarmby, has developed an equation for estimating car bus modal split for work trips to central London: pr (c / z ) = 2.26e1.04 ( z 0.431) 1 + 2.26e1.04 ( z 0.431)

pr(c/z)= the probability of choosing the car mode- given that the travel disutility is z. The disutility measure was developed as a function of differences in total travel time, excess travel time, costs and income related variables. Talvitie model: pr (m = 1 / ij ) = e z + ln( x / y ) 1 + e z + ln( x / y )

pr (m = 2 / ij ) =

1 1+ e
z + ln( x / y )

z is assumed to be normally distributed. Where Pr(m/ij) = the probability that an individual will use mode m given that the trip is between zones i and j. x, y = The a priori probabilities of membership in groups m=1 and m=2 respectively.
Probit Analysis

The basic premise is that as choice trip makers are subjected to changing magnitudes of relative trip costs, the proportion of trip makers that respond by choosing a particular mode of transport will follow a linear relation. Lave has developed the following equation for estimating the probability of bus-car modal patronage for Chicago area. Y = 2.08 + 0.00759kWT + 0.0186c 0.0254 IDC c + 0.0255 A R2 = 0.379

34

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay Where Y= binary variable with positive magnitudes denoting transit riders and negative magnitudes denoting car riders.
kWT = time difference between modes multiplied by the tripmakers wage rate

and his marginal preference for leisure time. IDCc = a binary valued comfort variable multiplied by income and trip distance.
Logit Analysis:

Stopher model.

pr ( m = 1 / ij ) =
pr (m = 2 / ij ) = where

e
1

* zij * zij

1+ e
* zij

1+ e

zij* = some function of the generalized costs of travel by modes m=1 and m=2

Two Stage Modal Split Model

Vandertol et al. have developed a simple two stage model which recognizes explicitly the existence of both captive and of choice transit riders. The model first identifies both the production and attraction trip ends of transit captives and choice transit riders separately. The two groups of trip makers are then distributed from origin to destinations. The choice transit riders are then split between transit and car according to a choice modal split model, which reflects the relative characteristics of the trip by transit and the trip by car. In most cities, the transit captive is severely restricted in the choice of both household and employment locations. Studies in a number of cities have shown that the trip ends of the transit captives tend to be clustered in zones that are well served by public transport. The challenge is to develop is to formulate a technique that uses information normally available in urban areas. Zonal work trip productions disaggregated by the captive and the choice transit riders may be estimated from Pi q = hitpq

35

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay Where Pi q = no of work trips produced in zone i by type q trip makers hi = the no of households in zone i tpq = work trip production rate for trip maker group q which is a function of economic status of a zone and the average no of employees per household. The work trips attracted to each zone j by trip maker type q may be estimated from ajq= [prcq] [rct] [etj] Where ajq= the no of work trips of type q trip maker attracted to zone j [prcq] = A row vector of the probability of the trip maker type q being in occupation category type c. [rct] = A c*t matrix of the probabilities of an occupation category type c within an industry type t. [etj] = a t*j matrix of the no of jobs within each industry type t in each zone j.

2.4 TRIP DISTRIBUTION

A trip distribution model produces a new origin-destination trip matrix to reflect new trips in the future made by population, employment and other demographic changes so as to reflect changes in people's choice of destination. They are used to forecast the origindestination pattern of travel into the future and produce a trip matrix, which can be assigned in an assignment model of put into a mode choice model. The trip matrix can change as a result of improvements in the transport system or as a result of new developments, shops, offices etc and the distribution model seeks to model these effects so as to produce a new trip matrix for the future travel situation. Trip distribution models connect the trip origins and destination estimated by the trip generation models to create estimated trips. Different trip distribution models are developed for each of the trip purposes for which trip generation has been estimated.Various techniques developed for trip distribution modeling are Growth Factor Models Synthetic Models

36

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Growth Factor Models

Uniform Factor Method Average Factor Method Detroit Method Fratar Method Furness Method Furness Time-function Iteration

Synthetic Models

Gravity/Spatial Interaction Models Opportunity Models Regression/Econometric Models Optimization Models

2.4.1 Growth Factor Models

1. Uniform Growth Factor Model A single growth factor for the entire area under study is calculated by dividing the future number of trip ends for the horizon year by trip ends in the base year. The future trips between zones I and j are then calculated by applying the uniform factor to the base year trips between zones i and j.

I
100 3

II 25
250 4

50 25 75
300
1

150

200 IV

400 2

III

37

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

The origin and destination matrix can used to represent the given network. I II III IV I 0 25 50 25 II 25 0 150 75 III 50 150 200 IV 25 75 200 From the above origin-destination matrix, we can the following in matrix form.
I I II III IV Tj II 25 III 50 150 IV 25 75 200 ti 100 250 400 300 1050 Fi 300 1000 300 300 Ti 300 1000 800 300 2400 Tcal 229 571 914 686 2400

25 50 25 300

150 75 1000

200 800

300

Uniform growth factor F(1) = 2400/1050 = 2.286 Tcal is tabulated using the calculated growth factor in the last column. Uniform growth factor F(2) = 2400/1050 = 2.286. Since the growth factors remain equal, there is no need of further iteration There are some drawbacks of using this method. These are: 1. The assumption of a uniform growth rate for the entire study area is not correct as the growth factor varies across zones. 2. The land use pattern changes with time, but not uniformly as assumed. Hence growth factor changes with time.
2. Average Growth Factor Model:In this method, the growth factor represents the average growth associated with both the origin and destination zones. If Fi and Fj are the growth factors for the zones I and j respectively, then: Where Fi = Ti/ti and Fj = Tj/tj Tij = tij(Fi + Fj) 2 After the first distribution, it may be found that the sum of the trips from zones are not equal to the projected trip ends for the respective zones. This discrepancy has then to be removed by successive iterations as: Fi' = Ti/ti' and Fj' =Tj/t'j where ti' and tj' are the generation and attraction of zone I and j respectively obtained from the first stage of distribution. This can be illustrated using the network given Above

38

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Iteration 1: -T11 =0(3+3)/2 =0 I I II III IV tj' Tj Fj' II 87 III 125 450

T12=25(3+4)/2=87 and so on
IV 50 187 300 ti' 262 725 875 538 2400 Ti 300 1000 800 300 Fi' 1.145 1.379 0.914 0.5576

88 125 50 263 300 1.145

450 188 725 1000 1.379

300 875 800 0.914

537 300 0.557 6

2400

Iteration 2: -The second iteration can be done in tabular form also. I I II III IV tj'' Tj II 110 II 129 516 IV 43 181 221 ti'' 282 808 866 444 Ti 300 1000 800 300 Fi'' 1.064 1.2376 0.9238 0.5576 0.6756

445 300 0.675 1.064 1.2376 0.9238 6 Fj'' Iteration 3: -The third iteration can be done in tabular form also. I II III IV ti''' 128 128 37 293 I 128 557 174 859 II 128 557 177 862 III 36 174 176 386 IV 2923 859 861 386 tj''' 300 1000 800 300 Tj 1.024 1.164 0.929 0.777 Fj'''
Iteration 4: -The fourth iteration can be done in tabular form also. I I II III IV II 140 II 125 582 IV 33 169 151 ti'''' 298 891 858 353

111 129 42 282 300

516 182 808 1000

220 865 800

2400

Ti 300 1000 800 300

Fi''' 1.024 1.164 0.929 0.5576 0.777

2400

140 125 33

582 169

151

Ti 300 1000 800 300

Fi'''' 1.0067 1.1233 0.9324 0.5576

39

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

tj'''' Tj Fj''''

298 300 1.0067

891 1000 1.1233

858 800 0.9324

353 300 0.849 8

0.8498 2400

Iteration 5: -The fifth iteration can be done in tabular form also. I I II III IV tj''''' Tj Fj''''' II 149 II 121 599 IV 31 166 ti''''' 301 914 Ti 300 1000 Fi''''' 0.99 1.09 0.9367 7 0.906

149 121 31 301 300 0.99 599 166 914 1000 1.09

134 134 854 800 0.93677 331 300 0.906

854 331

800 300 2400

Since the results obtained from the two successive iterations give approximately equal growth factors, I can stop the iteration. The disadvantages of this model are: 1.The factors do not have real significance. 2.Large number of iterations is required.

3. Detroit Model: This method is an improved version of the average growth factor method and takes into account the growth facor for the zones and average growth factor for the entire study area. where Fi =Ti/ti and Fj = Tj/tj and F= Total Tij/Total tij Tij = tij*FiFj F Iteration 1: -F1 = 2400/1050 = 2.2857 I I II III IV tj' Tj Fj' II 131 III 131 525 IV 33 131 175 ti' 295 787 831 339 2252 Ti 300 1000 800 300 Fi' 1.017 1.27 0.9627 0.885

131 131 33 295 300 1.017

525 131 787 1000 1.27

175 831 800 0.9627

339 300 0.885

2400

40

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Iteration 2: -F2=2400/2252=1.066 I I II III IV tj'' Tj Fj'' II 159 II 120 602 IV 28 138 140 ti'' 307 899 862 306 2374 Ti 300 1000 800 300 Fi'' 0.972 1.1124 0.925 0.9804

159 120 28 307 300 0.972

602 138 899 1000 1.1124

140 862 800 0.925

306 300 0.980 4

2400

Iteration 3: -F3=2400/2374=1.010952 I I II III IV tj''' Tj Fj''' II 171 III 108 615 IV 26 149 126 ti''' 305 935 849 301 2390 Ti 300 1000 800 300 Fi''' 0.9836 1.0695 0.9423 0.9967

171 108 26 305 300 0.9836

615 149 935 1000 1.0695

126 849 800 0.9423

301 300 0.996 7

2400

Iteration 4: -F4=2400/2390=1.0042 I I II III IV tj'''' Tj Fj'''' II 179 III 100 617 IV 25 158 118 ti'''' 304 954 835 301 2394 Ti 300 1000 800 300 Fi'''' 0.9868 1.0452 0.9581 0.9967

179 100 25 304 300 0.9868

617 158 954 1000 1.0452

118 835 800 0.9581

301 300 0.996 7

2400

Iteration 5: -F5=2400/2394=1.0025 I I II 185 III 94 IV 25 ti''''' 304 Ti 300 Fi''''' 0.9868

41

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

II III IV tj''''' Tj Fj'''''

185 94 24 304 300 0.9868

618 618 165 968 1000 1.03305 112 824 800 0.9709

165 112 302 300 0.993 4

968 824 302 2398

1000 800 300 2400

1.03305 0.9709 0.9934

since the ratio is approximately equal to 1, I can stop the iteration.

4. Fratar Method The total trips emanating from a zone are distributed to the interzonal movements and according to the relative attraction of each movement, locational factors for each zone are calculated. Then Tij=tij*Fi*Fj(Li+Lj) 2 The location factor values are computed below for the first iteration: L1= 100 0.444 25*1+50*2+25*4 L2= 250 0.556 25*3+150*2+75*1 L3= 400 0.421 150*4+50*3+200*1

L4= 25*3+75*4+200*2 Iteration 1: I I 150 II 130 III 36 IV 316 tj' 300 Tj

Fj' Lj'

300

0.387

II 150

II 130 586

586 161 897 1000 1.1149 1.14

IV 36 161 186

186 904 800 0.8849 0.9792

0.949 1.018

385 300 0.779 2 1.013 1

ti' 316 897 904 385 2502

Ti 300 1000 800 300

Fi' 0.949 1.1149 0.8849 0.7792

Li' 1.018 1.14 0.9792 1.0131

2400

Iteration 2: -The values in the table above can be used for this iteration. I II II IV ti'' Ti Fi'' Li''

42

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

I II III IV tj'' Tj Fj'' Lj''

171 171 109 27 307 300 0.9772 0.9829 612 150 933 1000 1.0718 1.048

109 612 129 850 800 0.9411 0.9562

27 150 129 306 300

0.9804 0.9917

307 933 850 306 2396

300 1000 800 300 2400

0.9772 1.0718 0.9411 0.9804

0.9829 1.048 0.9562 0.9917

Iteration 3: -The values in the table above can be used for this iteration. I I II III IV tj''' Tj Fj''' Lj''' II 182 II 97 619 IV 25 161 116 ti''' 304 962 832 302 2400 Ti 300 1000 800 300 Fi''' 0.9868 1.0395 0.9411 0.9934 Li'''
0.9893 1.0292 0.9737 0.9948

182 97 25 304 300 0.9868 0.9893

619 161 962 1000 1.0395 1.0292

116 832 800 0.9615 0.9737

302 300 0.993 4 0.994 8

2400

The main drawbacks for this method are: 1. It is tedious for even moderate sized problems 2. It does not take into account the effect of changes in accessibility of the study area. 5. Furness Method: This method estimates the future traffic originating and terminating at each zone and hence yields the origin growth factor and destination growth factors for each zone. The traffic movements are made to agree alternatively. Tij=tij*Ti/Total Tj and then Tij' = Tij*Ti/Total Tj' and Tij''=Tij*Ti/Total Tj''
Iteration 1: -Multiplying by the origin growth factors I I II III IV tj' II 75 II 150 600 IV 75 300 400 ti 300 1000 800 300 2400 Ti 300 1000 800 300

100 100 25 225

300 75 450 43

200 950

775

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Tj Fj'

300 1.333

1000 2.222

800 0.8421

300 0.387 1

2400

Iteration 2: -Multiplying by the destination growth factors I II 167 II 126 506 IV 29 116 155 ti'' 322 755 955 368 2400 Ti 300 1000 800 300 Fi'' 0.9317 1.3245 0.8377 0.5152

I 133 II 133 667 III 34 166 168 IV 300 1000 800 tj'' 300 1000 800 Tj Iteration 3: -Multiplying by the origin growth factors I I II III IV tj''' Tj Fj''' II 156 II 117 670

300 300

2400

176 111 28 315 300 0.9524

559 135 850 1000 1.1765

IV 27 154 130

137 924 800 0.866

311 300 0.964 6

ti''' 300 1000 800 300 2400

Ti 300 1000 800 300

2400

Iteration 4: -Multiplying by the destination growth factors I II II IV 183 101 26 I 168 580 149 II 106 858 125 III 26 158 119 IV 300 1000 800 300 tj'''' 300 1000 800 300 Tj Iteration 5: -Multiplying by the origin growth factors I I II III II 177 III 98 647 IV 25 166 113

ti'''' 310 897 889 304 2400

Ti 300 1000 800 300

Fi'''' 0.9677 1.115 0.899 0.9868

2400

187 95

592

ti''''' 300 1000 800

Ti 300 1000 800

44

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

IV tj''''' Tj Fj'''''

26 308 300 0.974

157 926 1000 1.08

117 862 800 0.928

304 300 0.986 8

300 2400

300 2400

Successive iterations are done until growth factors approach unity.

6.Time Function Method: This method assumes that the trip distance is influenced by the journey times and row and column totals the trip ends.

I I II III IV
Current Destination

II 25

III 50 150 200 400

IV 25 75 200 300

25 50 25 100

150 75 250

Current Origin 100 250 400 300

I I II III IV
Current Destination Ultimate Destination

II 1

III 1 1 1 3 400

IV 1 1 1 3 300

Current Origin 3 3 3 3

Ultimate Origin

1 1 1 3 100

1 1 3 250

100 250 400 300

Successive iterations are performed by alternately matching with the ultimate origins and destinations and finding the adjustment factors for column and row totals respectively. Iteration 1: I I II III IV 333 266 100 II 100 267 100 III 100 334 100 IV 100 333 267 Current Origin 300 1000 800 300
Ultimate Origin

300 1000 800 300

45

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Current Destination Ultimate Destination Adjustment Factor

699 300 0.629

467 1000 2.141

534 800 1.498

700 300 0.428 6

2400 2400

Iteration 2: -

I I II III IV
Current Destination Ultimate Destination

II 214

III 150 500 150 800 800

IV 43 143 114 300 300

Current Origin

Ultimate Origin

Adjustment Factor

143 114 43 300 300

572 214 1000 1000

3 3 3 3 2400

300 1000 800 300

0.7371 1.2722 1 0.7371

2400

2.4.2 Synthetic Models Gravity Model

The trip distribution models found most often in practice today are "gravity models," so named because of their basis in Newtons law. The gravity model assumes that the trips produced at an origin and attracted to a destination are directly proportional to the total trip productions at the origin and the total attractions at the destination. The calibrating term or "friction factor" (F) represents the reluctance or impedance of persons to make trips of various duration or distances. The general friction factor indicates that as travel times increase, travelers are increasingly less likely to make trips of such lengths. Calibration of the gravity model involves adjusting the friction factor. The socioeconomic adjustment factor is an adjustment factor for individual trip interchanges. An important consideration in developing the gravity model is "balancing" productions and attractions. Balancing means that the total productions and attractions for a study area are equal.

46

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay Standard form of gravity model

Where: Tij=trips produced at I and attracted at j Pi = total trip production at I Aj = total trip attraction at j F ij = a calibration term for interchange ij, (friction factor) or travel time factor ( F ij =C/tijn) C=calibration factor for the friction factor Kij = a socioeconomic adjustment factor for interchange ij I=origin zone n = number of zones Before the gravity model can be used for prediction of future travel demand, it must be calibrated. Calibration is accomplished by adjusting the various factors within the gravity model until the model can duplicate a known base years trip distribution. For example, if you knew the trip distribution for the current year, you would adjust the gravity model so that it resulted in the same trip distribution as was measured for the current year.
CALIBRATION OF GRAVITY MODEL

The most widely used technique for calibrating the form of the gravity model defined in equation Tij = Pi {(aijfij)/

a jfij}
j

is that developed by Bureau of Public Roads. The purpose of the calibration procedure is to establish the relationship between fij and zij for base year conditions. This function is then used along with equation to develop a trip interchange matrix that satisfies the constraint equations. The Bureau of Public Roads calibration procedure is directed

47

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

toward the development of a travel time factor function, which is assumed to be an area wide polynomial function of interzonal travel times. Figure below shows the sequence of activities involved in the calibration of the gravity model .the first step involves the estimation if inter centroid travel time for each centroid pair. It is suggested that the gravity model simulated and observed trip-length frequency distributions should exhibit the following two characteristics:

(1) The shape and position of both curves should be relatively close to one another when compared visually. (2) The differences between the average trip lengths should be within 3 percent.

If the trip length frequency distribution produced by the gravity model does not meet these criteria, then a new set of travel factors may be estimated from the following expression:

f = f * (OD%)/(GM%) Where f = the travel time factor for a given travel time to be used in next iteration. f = the travel factor used in the calibration just completed. OD% = the percentage of total trips occurring for a given travel time observed in the travel survey. GM%= the percentage of total trips occurring for a given travel time observed in the simulated by the gravity model.

48

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

CALIBRATED PARAMETERS

Observed Pj + Aj by Trip Type

Interzonal Skim Trees

Assumed initial fij Vs dij

Calculate Tij Matrix

Calc Trip length Freq Distbn

Compare Sim and Obs Distrbn

New

fij Vs dij relation

fij Vs dij for trip type

Calc Balanced attraction A* j

Calculate New Tij with Aj

Calculate Kij factors

Kij Factors

49

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

The final phase of BPR calibration is to calculate zone to zone adjustment factors kij.These factors are calculated from the following expressions: kij = rij [(1-xij)/(1-xirij)] where kij = the adjustment factor to be applied to movements between zones I and j. rij = the ratio tij (o-d survey)/tij (gravity model) xij = the ratio tij (o-d survey)/pi The final gravity model simulated trip interchange matrix is given by tij = pi [ (aj*fijkij)/

a j*fijkij ]

An horizon year trip interchange matrix is calculated from the given equation with the following inputs: (1) The horizon year trip production and trip attraction rates, (2) The horizon year intercentroids skim trees. (3) The base year travel time function. (4) The kij magnitudes that are expected to hold for the horizon year.
Limitation:

The limitation of procedure described is it requires that two criteria be satisfied by a base year calibration. These two criteria are: agreement between observed and simulated trip length constraint equation .A principal difficulty wit this calibration procedure is that the travel time factor function and associated trip length frequency distribution are assumed to be constant for each zone of a study area.

50

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Calibrate the Gravity Model for 4-network problem

1
P1=500

5 min 3
A3 = 300

10 min 4
A3=800

10 min 5 min 2
P2 = 600

t13 200 5
Iteration 1 Travel tij Time 5 700

t14 300 10

t23 100 10

t34 500 5

fij = t (GM) ( 1/ dij)2 ij 0.04 849 0.01 251

fij = f X tij (OD)/ tij (GM) 0.032 0.015

10

400

Iteration 2 5 700

.032 .015

735 365

.0304 .016

10

400

51

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Iteration 1:

Calculation Attraction Origin 1 tt fij =( 1/ dij)2 aj x fij (aijfij)/

Destination 3 4

5 0.04 12 0.6 300

10 0.01 8 0.4 200 20 1 500

a j fij
j

Pi aijfij/

a j fij
j

Origin 2 tt fij =( 1/ dij)2 aj x fij (aijfij)/ 10 0.01 3 5 0.04 32 0.915 549 35 1 600

a j fij
j

0.085 51

Pi aijfij/

a j fij
j

Iteration 2
Calculation Attraction Origin 1 Tt
fij =( 1/ dij)2

Destination 3 4

5 .032

10 .015

52

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

aj x fij

9.6

12 0.55 275

21.6 1 500

(aijfij)/

a j fij
j

0.45 225

Pi aijfij/

a j fij
j

Origin 2 tt
fij =( 1/ dij)2 aj x fij

10 0.015 4.5

5 0.032 25.6 0.85 510 30 1 600

(aijfij)/

a j fij
j

0.15 90

Pi aijfij/

a j fij
j

LOWS METHOD
Basic Concept: In this model volumes are determined one link at a time, primarily as a function of the

relative probability that trips would use one link in preference to another link. First trip probabilities are determined for every combination of origin and destination zones in the area. The probability of a trip between zone I and zone j can be linked to the gravitational pull of two masses and the distance separating them as
mi * m j
2 d ij

Considering home work trip if mass at home end as employment Ej ,then the trip Pi E j probability becomes m t ij Input Information The Lows model needs the following information

53

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

1. Pattern and intensities of land use development now and as anticipated in the future. This should include population and employment by zone etc but there will be of no values unless reasonable estimates of future patterns and intensities of land estimates of future patterns and intensities of land use development can be made in similar detail. 2. Transportation network characteristics including network configuration, link speeds etc. 3. Representative traffic volumes from ground count throughout the network. 4. Volume and Patterns of trips with one or both ends outside area under study.
Formation and Use of the Model Current external volumes:

Current external trip data gathered in the road side interview survey are first assigned to the existing network to produce estimates of current external volumes throughout the network.
Current Internal Volume:

These volumes on links throughout the network are computed by subtracting the assigned external traffic volumes from the corresponding ground counts.
Internal Volume Forecasting Model:

Inter zonal trip opportunity matrices of the form Ai and Bj are developed Where A and B are parameters that are logically related to trip productions and attractions. Using Travel time as the measure of separation the friction factor can be expressed as 1/ tm ij The product Fnij= Ai Bj / tm ij is called inter zonal trip probability matrix. The probability matrices are then assigned separately to the current network just as if they were trips. Multiple regression techniques are used to develop equation of the following form

V = a1 + b1 F1 + b2 F2 + b3 F3 + ...... + bn Fn

54

CE -751, SLD, Class Notes, Fall 2006, IIT Bombay

Where V is the internal traffic volume on a link a and b are constants Fn is the trip probability factor volume as assigned to that link.
Future Internal Volume:

Future zonal socioeconomic data are used to develop future trip opportunity matrices and future friction factors from the future network to be tested are applied to the trip opportunity matrices to produce future trips. Advantages of Lows Method They are easily understood and applied requiring only as inventory of present day trip origins and destinations and estimations of simple growth factors. The simple process of iteration quickly produces a balance between postulated and computed trip ends. They are flexible in application and can be used to distribute trips by different modes for different purposes at different times of the day and can be applied to directional flows. They have been well tested and have been found to be accurate when applied to areas where the pattern and density of development is stable. Disadvantages of Lows Method They cannot be used to predict travel patterns in areas where significant changes in land use are likely to come and the assumption that the present day travel resistant factors will remain constant into the future is fundamentally weak. These models cannot satisfy the requirements modern urban transportation studies, which are usually designed to cater for conditions of continual and rapid changes in the pattern of development and the way of life of population generally.

55