A trip on the public transport network involves walking to the bus stop, travelling on the bus to the bus stop closest to your destination, and then walking to your final destination. It may also include transferring to another bus, or other modes of public transport such as ferries and trains. The following description refers to creating a public transport network for buses, although a similar approach is applicable for other modes.
The public transport network model was constructed using the following information:
· walking network
· location of bus stops
· location of bus routes
· bus frequency details
· bus stopping details (which bus stops are used by which bus routes).
With the exception of air bridges in the CBD, everything in the walking network is on the same (conceptual) grade or level. This means whenever two edges on the network cross, a person can transfer from one to the other, ie they can move from one footpath to another wherever they meet. This is not the case with public transport. It is not always possible to transfer between buses wherever their routes cross. Instead one must travel to the nearest bus stop, alight the first bus, wait for the next bus to arrive and then continue one’s journey on the second bus.
In order to enforce this logic, each bus route was added to the network dataset on a separate level. This meant the different bus routes passed over each other without connecting which enforced the requirement to stay on the bus between stops.
To model the activity of boarding and alighting a bus at a bus stop, connecting links were drawn between the walking network (on level 1) and each bus route (on different levels) that stopped at that bus stop to pick up or drop off passengers.
A schematic diagram of the public transport network is illustrated in figure 14.3, showing the walking network on level 1 and each bus route on successive levels. Links were then made between the walking network on level 1 and the relevant levels for the bus routes which stopped at that location.
These links for boarding the bus were attributed with a delay to model the time spent waiting for a bus to arrive. This value was set at half the timetabled frequency of the bus or a maximum of 7.5 minutes.
The value was capped at 7.5 minutes (450 seconds) on the basis a passenger would not leave their house for the bus stop until close to the anticipated time for the bus to arrive. No delay was assigned to these links for alighting the bus. All of these connecting links were digitised in the same direction, ie from the walking network to the bus route. This meant the delay could be applied to just one column in the table, ie it only needed to be added to the (FT) digitised direction and not the reverse (TF) direction.
Note: if a bus turned around at the end of its route with passengers still onboard, a single link was required to connect the levels of the inbound and outbound bus routes.
Each edge of the network included the speed at which it was possible to travel along that edge (path) stored as an attribute of the edge. For the edges which were part of the walking network and the links for boarding and alighting public transport, this speed was set to an average walking speed of 1.3m per second. This speed should be varied depending on the ability of the target user being assessed, ie slower speeds should be applied for less able users and faster speeds for more able or purposeful users (such as commuters).
For the edges which formed part of a bus route, the average speed of the bus along the bus route was calculated from the scheduled timetable. An average speed was calculated based on the travel time between stops recorded on the timetable and the distance along the bus route between these two stops. This average speed was then applied to all the edges that comprised that section of the bus route. This was repeated for the entire length of each bus route in the public transport network model.
In the UK it has been observed that bus timetables are not regular throughout the day. It is very common for express buses to offer a 20-minute journey in peak hours to a town centre and for the rest of the day take more than twice this time for the same journey. Therefore the question arises whether the journey time used in the accessibility modelling should be 20 minutes or 40+ minutes. Between 2006 and 2009 the UK analysis assumed a median journey time of six modelled time periods but this proved to be unstable so since 2009, 23 modelled time periods have been used and a probabilistic method is used to calculate the journey time based on the probability of a journey time being chosen. Shoppers for example might choose to travel at a time when a faster journey time is available but people travelling to health appointments have less choice of journey time so the journey time is an average of fast and slow speeds. Some travel needs are largely made off peak, such as attending evening education classes, so would be constrained to slower journey times.
The New Zealand public transport timetable data should allow a similar sort of approach but undertaking 23 model runs for each trip purpose and then averaging the results is a substantial task involving considerable resources.
The cost of travelling along an edge in the network was modelled as the time (in seconds) taken to traverse the edge. For the edges on the walking network, while onboard a bus and when alighting a bus, this cost was calculated as:
Edge traversal time = edge length/edge traversal speed
When boarding a bus, the delay due to waiting for the bus to arrive was added to the edge traversal time to give the following equation:
Edge traversal time = (edge length/edge traversal speed) + waiting delay
The edge traversal time in seconds was then stored in a field called ‘Seconds’ and used to perform travel time calculations across the public transport network.