Appendix JTravel Demand Modelling Process J.1 Trip Generation

The first step in travel forecasting is trip generation. In this step, information from land use, population, and economic forecasts are used to estimate how many trips will be made to and from each zone.

Types of Most Commonly Used Trip Generation Models

Regression models: equations such as:

Total trips = 1.1*population,

Home-based work trips = 1.5*total employment, etc;

Cross-classification tables based on: household size, number of workers, income class, vehicle ownership, etc.;

Special generator models: used for generators such as airports, colleges, amusement parks, green spaces, etc.;

Separate rates for peak and off-peak trips.

Assumptions and limitations:

Pedestrian and bicycle trips are usually excluded from trip generation models.

Commercial vehicles need to be generated separately.

Care should be taken before interchanging vehicle trip rates with person trip rates.

Compare trips per household to the regional average. There should be a valid explanation if numbers are too high or too low.

Review trips per household by purpose.

Ensure that trip productions and trip attractions balance well. Check trip rates and adjust if they do not balance

Create GIS color-coded maps of productions and attractions to review for errors. Use special generators for facilities such as schools, hospitals, malls, and recreational facilities such as parks, playgrounds, etc.

Make model external trips and visitor trips using external cordon and visitor surveys.

Make sure Commercial Vehicle and Freight Travel is included in the model

The trip distribution model estimates number of trips travelled from one zone to another. Trip generation only finds the number of trips that begin or end at a particular zone. The process of trip distribution links the trip ends to form an origin-destination pattern. Trip distribution is used to represent the process of destination choice (i.e. "I need to go shopping, but where should I go to meet my shopping needs?").

The most commonly used procedure for trip distribution is called the gravity model. The gravity model takes the trips produced at one zone and distributes to other zones based on both the size of the other zones (as measured by their trip attractions) and the composite cost (travel time, distance or cost for all modes available) to other zones. A zone with many trip attractions (say, a large shopping center) will

receive a greater number of distributed trips than one with few attractions (a small shopping center). The mathematical formula for a gravity model is shown below:

Where,

$T _ { j } { \overline { { \overline { { \phantom { . } } } } } } ^ { \mathrm { { { } { { { { { \mathrm { { { \scriptsize ~ } } } } } } } } } } }$ the number of trips from zone i to zone $j ,$

$P _ { i }$ = the number of trip productions in zone

$A _ { j } { = }$ the number of trip attractions in zone j,

$F _ { i j } { = }$ the friction factor relating the spatial separation between zone i and zone j, and

$K _ { i j } { = }$ an optional trip-distribution adjustment factor for interchanges between zone i and zone j.

Distance to possible destinations is the other factor used in the gravity model. The number of trips to a given destination decreases as the distance to the destination increases (it is inversely proportional). This is represented in a factor called friction factor. It is inversely proportional to the distance or travel time between two zones. Friction factors are used to calibrate trip lengths from the model to observed data. The most common formulation of friction factors is through a gamma function as shown below:

Where,

$\mathsf { F } _ { \widehat { \mathbb { Y } } }$ = the friction factor between zones i and j,

a, b, and c = model coefficients; both b and c should, in most cases, be negative; a is a scaling factor and can be varied without changing the distribution,

$t _ { \mathrm { i } } { = }$ the travel time between zones i and j, and

e = the base of the natural logarithms.

Validation and Reasonableness Checking for trip distribution

Compare trip length frequency from model and survey data.

Do this for every purpose

Do this for distance and time

If different, adjust friction factors in the model

Maximum of 5-10% difference in observed and modelled.

Compare trip patterns from the model to data from O-D Surveys.

Aggregate data into districts and compare trips

Look at north-south and east-west trips

Mode choice is one of the most critical parts of the demand modelling process. It is the step where trips between a given origin and destination are split into trips using available modes such as two-wheeler, car, private vehicle, public transport, etc. Calculations are conducted that compare the attractiveness of travel by different modes to determine their relative usage. All proposals to improve public transport or to change the ease of using private vehicles are passed through the mode split/private vehicle occupancy process as part of their assessment and evaluation. It is important to understand what factors are used and how the process is conducted in order to plan, design, and implement new systems of transportation. The most common mathematical formulation of a mode choice model is a logit model and is as follows:

Where,

$\mathsf { P } _ { \mathrm { i } } =$ the probability of a traveller choosing mode i,

$\mathbf { u } _ { \mathrm { i } } = \mathbf { a }$ linear function of the attributes of mode i that describe its attractiveness, also known as the utility of mode i, and

$\sum _ { i = 1 } ^ { k } e ^ { u i }$ = the summation of the linear functions of the attributes of all the alternatives, k, for which a choice is available

The utility function is calculated based on the following equation:

Where,

IVTT_i = the in-vehicle travel times for mode i,

OVTT_i = set of variables measuring the out-of-vehicle travel times for mode i—walk, wait, and transfer times—may all be kept separate or combined, depending on the calibrated structure of the model,

COST_i = the cost of mode i,

a_i = mode-specific coefficient (constant) to account for mode bias not measurable with the level of service variables,

b_i= coefficient for the IVTT variables of mode

c_i = a set of coefficients for OVTT variables of mode i, and

d_i= coefficient for COST variable of mode i

Mode split is done by a comparison of the "disutility" of travel between two points for the different modes that are available.

"Disutility" is a term used to represent a combination of the travel time, cost, and convenience of a mode between an origin and a destination.

Travel time is divided into two components: in-vehicle time to represent the time when a traveller is actually in a vehicle and moving, and out-of-vehicle time, which includes time spent travelling outside of the vehicle (time to walk to and from bus/rail stops, waiting time, transfer time).

Out-of-vehicle time is used to represent "inconvenience" and the coefficient for out-of-vehicle time is typically several times larger than for in-vehicle time, reflecting the fact that travellers do not like to wait or walk long distances to their destinations. The size of the multiplier will be different depending upon the purpose of the trip.

For public transport trips, the cost of the trip is usually measured as the average public transport fare for that trip, while for private vehicle trips, cost is found by adding the parking cost to the length of the trip and multiplying by cost per kilometer. Private vehicle cost is based on a "perceived" cost per kilometer (on the order of 5-10 cents/kilometer), which only includes fuel and oil costs and does not include ownership, insurance, maintenance, and other fixed costs (total costs of private vehicle travel are much higher).

Disutility equations also contain a "modal constant", represents other characteristics or travel modes that are not explicitly captured by the variables in the model, but that may influence the choice of mode (such as a difference in comfort between public transport and private vehicles). A model will have n-1 constants with ‘n’ being the number of modes.

Once disutilities are estimated for the various choices between an origin and a destination, the trips are split among various modes based on the relative differences between disutilities. A large advantage will mean a high percentage for that mode.

All model coefficients are estimated using actual traveller data.

Factoring of vehicle trips - In this model, fixed factors are used to factor total trips into trips by mode. This is based on fixed data from surveys. This is not recommended for long-term projects with high investment

Binomial logit model - In this model, only two modes are considered: private vehicle and public transport. This is a simpler model and should be used when short on time and data.

Multinomial logit model - This is the most common type of model used to compare a number of modes. Variations of the model include nested logit and hierarchical logit, which splits total trips first into public and private vehicles, and then into categories such as two- and four-wheelers, bus, rail, etc.

Table J.1:Factors Affecting Mode Choice

Source: Development of Training Material under Sustainable Urban Transport Project, Reference Guide Volume 2 Demand Assessment, MoUD

Compare observed modal split to estimated modal split

Do this for every purpose

– Maximum of 5-10% difference in observed and modelled.

A typical range for value of in-vehicle time coefficient is between -0.015 and -0.02.

A typical range for value of out-of-vehicle time coefficient is 2- 3 times the in-vehicle coefficient.

Constant terms should not be too high or low.

Perform sensitivity tests to make sure model is stable.

Increase/decrease travel times/fares of highway and PT

Change socioeconomic data

The first three steps of modelling generate total person trips in the region. Before assigning them to road network, person trips need to be converted into vehicle trips. The first step in this process is to split private vehicles trips into private vehicle driver and private vehicle passenger trips. Private vehicle occupancy analysis is often a highly simplified process that uses fixed private vehicle occupancy rates for a given trip purpose or for household size and private vehicle ownership categories.

Trips are then converted into an origin-destination format for conducting daily assignments. Traffic varies considerably throughout the day and during the week. The travel demand forecasts are made on a daily basis for a typical weekday and then converted into peak-hour conditions. Daily trips are multiplied by an "hour adjustment factor", for example, 10%, to convert them into peak-hour trips. The number assumed for this factor is very critical. A small variation, say, plus or minus one percent, will make a significant difference in the level of congestion that would be forecast on a network. Most models are unable to represent how travellers tend to cope with congestion by changing the time they make their trips, although advanced travel demand models and activity models represent time of travel explicitly.

Once trips have been split into highway and public transport trips, the specific route that they use to travel from their origin to their destination must be found. These trips are then assigned to that route in the step called traffic assignment. Traffic assignment is conducted differently for highway trips and public transport trips.

The process first involves the calculation of the shortest path from each origin to all destinations (usually the minimum time path is used). Trips for each O-D pair are then assigned to the links in the minimum path, and the trips are added up for each link. The assigned trip volume is then compared to the capacity of the link to see whether it is congested. If a link is congested, the speed on the link needs to be reduced to result in a longer travel time on that link. When speeds and travel times are changed, the shortest path may change. Hence, the whole process must be repeated many times (iterated) until there is an equilibrium between travel demand and travel supply. Trips on congested links will be shifted to uncongested links until this equilibrium condition occurs. There are a variety of ways in which the calculations are done to reach network equilibrium. One way to get a feel for the accuracy of the models is to look at the resulting speeds on the network. These should be realistic after equilibrium.

Public transport trip assignment is done in a similar way, except that public transport headways are adjusted rather than travel times. Public transport headways (minutes between vehicles) affect the capacity of a public transport route. Low headways mean that there is more frequent service and a greater number of vehicles. Public transport supply and demand are also recalculated to reach equilibrium between supply and demand.

It is important to understand the concept of equilibrium. If a highway or public transport route is congested during peak hours, its excess trips will be shifted to alternative routes. If the alternative routes are also congested, the final results will show congestion over a wide part of the network. In the real world, this congestion will eventually dissipate overtime.

Another important step in assignment is the time of day analysis. Daily trip patterns need to be converted into peak-time period traffic. A key assumption needed is the portion of daily travel that occurs during the peak period. This is normally used as a constant, and conventional travel models have very limited capability to describe how travellers will shift their trips to less congested times of the day.

Validation and Reasonableness Checking for traffic assignment

The following summaries should be prepared:

Compare traffic volumes with observed counts.

Volumes by road class

Volumes by area type

– Volumes by screenlines

Volumes at external cordons

Observed vs. estimated speeds

Compare PT ridership from the model with observed ridership.

Ridership by route groups should be within +/- 15%

J.5 Model Forecasts

Forecasting is the process of using a fully calibrated and validated model to estimate future year traffic volumes and public transport ridership.

• Socioeconomic data, such as population, income class, and employment forecasts;

• List of approved highway and public transport projects and a list of projects to be evaluated.

Transportation forecasting processes should consider a broad range of alternatives. There can be alternative modes of transportation, alternative locations of different systems, alternative levels of capacity, or alternative policies. This would include the following:

A no-build alternative: This is basically the status quo with continuing maintenance and operation of the current transportation system. It may include substantial efforts to improve the efficiency and utilization of existing transportation systems. These are considered as an alternative to building new systems. Developing a no-build alternative is a must for evaluating other alternatives. Every new investment alternative has to be compared to the no-build alternative to see if the improvement justifies the cost.

Travel demand management policies: Transportation services require a broad range of policies to function. Travel demand management involves efforts to make the current system more efficient and to use techniques to reduce demand during critical periods. This is not unlike programs that utility companies have used to promote energy conservation among their customers. Some of the techniques that could be used include: use of priority techniques for high occupancy vehicles, parking regulation, efforts to shift when travel occurs, promotion of telecommuting, public transport service improvements, use of pricing techniques, etc.

Land use alternatives: Different land use patterns and policies can be used to affect travel demand and to affect the use of natural resources. These could include concentrated urban development patterns, increased suburban growth, or some combination.

Modal alternatives: Different modes of transportation such as highway, public transport, ride sharing, freight, etc. should be considered. Generally, transportation plans are concerned with picking the best combination of modes to deal with a particular problem. Single-mode plans such as a highway plan, public transport plan, etc. should not be done without first considering other modes. All options should be considered in order to develop a balanced transportation system.

Capacity changes: The capacity of each mode (number of lanes, how often public transport vehicles operate) can be varied in different alternatives. Highway capacity depends on many factors besides the number of lanes, such as intersection characteristics, traffic signal systems, and the characteristics of access along the highway.

Alternative locations: Proper location of public transport and highway facilities is an important part of their success. Facilities should be located to serve travel markets and to facilitate land development patterns that are good for the community.