SUPERNOVA – the evaluation tool

In late 2002, the OBB Austrian Federal Railways Passenger Transportation Department decided to develop a planning and evaluation tool based on a countrywide traffic-demand-model. The main aim was to evaluate the impact of new public transport services on passenger demand, hence SUPERNOVA: Simulationmodel on Use of Public Transport Systems for the Evaluation of Railway Network Optimising Variants.

SUPERNOVA was launched in October 2002 and transferred from project status to regular operation in Spring 2005. As the management was expecting the first results rather quickly, widespread modeling software was chosen and the input data was taken over from an existing model in the first stage of operation.

Software implementation and stage 1 of model development

Introduction of VISUM by the OBB

The model is based on the software VISUM by PTV AG, Karlsruhe. The main reason for this decision was the fact that VISUM had been developed and improved for many years based on the experience of a large number of users worldwide, with the aim of meeting the demands of a public transport company in the following years.

At the same time, the Austrian Traffic Demand Model (VMO), developed and maintained by the Ministry of Transport, Innovation and Technology (BMVIT), was taken over by OBB. When this model was developed between 1993 and 1998, OBB received the rights to use it by providing the BMVIT with passenger census data in order to calibrate the model properly. The following adaptations to the VMO appeared to be necessary.

Disaggregation of cells

In Austria, the cells (as traffic generating and attracting spatial units are called) are defined by the borders of the communities. Foreign cells are based on NUTS3-regions, with aggregation to larger units depending on the distance to Austria. For example, the UK is only one single cell whereas Bavaria is split into roughly 30 units. Finally the model consisted of 2,600 cells in total (2,400 Austrian communities and 200 units abroad).

Trains based on real-time schedule

The VMO included the simplified schedule of the train lines in Austria which was by far the evaluation of infrastructure investments carried out by the BMVIT. For a train operating company a much more detailed level was necessary leading to an automated solution. Via the HAFAS2VISUM-interface the real-time schedule was imported into VISUM, meaning that all 5,000 single trains running in Austria (including cross border traffic) on an average working day could be displayed in the model. Also, foreign trains were included as far as necessary to ensure that all the international traffic flows to, from and through Austria could be routed. Needless to say it was necessary to include all 1,900 train stations in Austria as a base for modeling the trains.

Access points to the network

To fill in the trips generated within a cell to the network (both modes – road and public transport), at least one access point per cell and mode has to be defined. As far as road traffic is concerned, this can be generated by an automated workflow provided in VISUM. For public transport, a manual definition (or at least a detailed check of the automatically generated access points) is recommended. In the case of more than one stop per cell, the most used station should be chosen.

Adaptation of parameters for traffic assignment

The VMO assigns the trips by public transport according to the travel time by summing up the time needed for accessing, waiting, riding and transferring during a journey. For the OBB other factors had to be added to this simplified approach.

In reality, the fare (based on the distance travelled) is influencing the assignment. A shorter route with a longer travel time is usually preferred by the users, as it is cheaper than the alternative. Neglecting this restriction, VISUM generates ‘backtrack effects’; The passenger takes a fast train between two hubs and rides back on a regional train along the same line to reach their final destination.

Also, the comfort of the rolling stock has to be considered as a factor in the parameters for traffic assignment. The necessary data was provided by passenger surveys containing the evaluation of the different types of cars by the customers. Although the influence of this comfort factor is rather restricted compared to travel time, effects of passengers preferring new rolling stock to old cars can be monitored in the model.

Update of the O-D-matrices and consideration of diurnal variation

The VMO contained origin-destination-matrices (O-D-matrices) for five different activities:

• Work
• Education
• Business
• Leisure
• Shopping and private purpose

As the matrices for the status-quo were based on the Austrian Census 1991 and therefore outdated at the time of introducing SUPERNOVA’s stage 1 in early 2003, an update was necessary. The O-D-matrices were calibrated by using the VISUM-add-on VStromFuzzy, based on passenger census data from 2002 provided by MOFAS, a modular passenger census system that had been developed at the OBB in 2000.

To ensure the correct allocation of passengers to the individual trains, the diurnal variation for the traffic demand of each activity had to be considered. Based on a countrywide survey carried out in 1995, these parameters were included in the model. The O-D-matrix for work had to be split in order to respect the different diurnal variations for office and homebound traffic. The same procedure was applied to education traffic.

Key figures from stage 1

After finishing the adaptations of the VMO by May 2003, which also marked the end of development stage 1, SUPERNOVA consisted of:

• 15,200 nodes (including 1,900 train stations)
• 33,00 links (including 6,500 rail tracks and 26,500 roads)
• 5,050 train lines (4,950 real-time individual trains and 100 simplified lines) of the timetable 2003
• 2,600 cells (2,400 in Austria and 200 abroad)
• 14 O-D-matrices (seven each for public transport and individual traffic)

By April 2005 when SUPERNOVA stage 1 was phased out, it had been used to evaluate more than 40 different projects consisting of about 150 variants. This equals an average of two evaluated variants per week.

Stage 2 of the model development

While operating SUPERNOVA’s stage 1 to evaluate current projects, the second stage of model development was launched in late 2003. The main focus was to cover the entire public transport system (especially bus operator Postbus, affiliated to the OBB Passenger Transport Department), making it necessary to have a much more detailed road network. When working on stage 2, a variety of technical problems had to be faced, caused by the volume and size of the data model which was well beyond anything known so far by PTV.

Establishment of a new network

Based on a digital road map by NAVTEQ used with onboard navigation systems, a new network had to be created. To reduce the amount of nodes (one million) and links (two million), a series of simplifications had to be applied, including lowest rank roads and deletion of nodes with only two adjacent links.

This was followed by the mapping of train stations and bus stops and snapping them on the links. In Austria, all stations available in the OBB timetable information system (including urban bus lines) were included; in foreign countries at least one station per cell was chosen.

The ‘simplified’ network consisted of 290,000 nodes (including 24,000 stations/stops) and 700,000 links.

Import of timetable information

A big challenge was to transfer 47,000 individual train and bus service trips from the timetable information system HAFAS to VISUM. As the interface provided by PTV was not able to handle such a vast number of service trips at once, the data had to be split and imported in several steps.

Definition of access points and hubs

In the first stage, usually only one train station per cell was defined as an access point. As there was now an additional large number of bus lines serving each cell, the stops for accessing the network had to be chosen strictly on a manual basis, with up to ten stops per cell depending on the size and public transport system of each cell. The choice had to respect not only the distance from the settlement area, but also the number of buses serving each stop.

To simulate a realistic behaviour of users concerning transfers, all hubs (which can include up to ten single stops) according to the timetable information system were implemented in the simulationmodel, including the official transfer time between each pair of stops.

Calculation of new O-D-matrices

Based on the Austrian Census 2001, new O-D-matrices for work and education commuters were generated. After assigning the new traffic demand to the network and comparing the result with the MOFAS passenger census data, it became obvious that the survey data was highly overestimated. So the decision was made to use the newly generated O-D-matrices of the Traffic Forecast 2025+ carried out on behalf of the BMVIT.

Assignment to the network and calibration

The final challenge was the assignment of the traffic demand on the public transport system. A choice of 47,000 single service trips meant many alternatives for each passenger and required a lot of calculation time and RAM. As Windows does not support more than 3GBs of RAM, the assignment was terminated after a while due to memory overflow.

To solve this problem, another simplification thin-out process had to be applied to the network, resulting in the deletion of all local streets, except where bus lines were running. This helped reduce the file size of the VISUM version from 600MB to 250Mb. Now, the RAM needed to decrease significantly, to result in a stable operation of the model.

Key figures of SUPERNOVA stage 2

When it was completed at the end of 2004, SUPERNOVA stage 2 had increased to the following dimensions:

• 110,000 nodes (including 24,000 train stations and bus stops)
• 250,000 links (including 15,000 rail tracks and 235,000 roads)
• 47,000 public transport lines based on the timetable information of an average working day
• 2,600 cells (2,400 in Austria and 200 abroad)
• 14 O-D-matrices (seven each for public transport and individual traffic)

SUPERNOVA in daily operation

Method of evaluation

Before focusing on the results generated by SUPERNOVA, the method on which the evaluation is based has to be described.

Generally, SUPERNOVA follows the classic four step algorithm of traffic models. The only difference to the common proceedings is that step one (trip generation), two (trip distribution) and three (transport mode choice) are not calculated within the model. In fact, the O-D-matrices for the status-quo come from an external database.

The O-D-matrices for the variants are calculated by VISEM/MUULI, which is also a product of PTV. Along with the traffic assignment, each mode of the matrix containing the generalised costs in each O-D-relation is calculated for the status-quo and the variant. By comparison in the modal-split-model the reaction of each activity group on increase or decrease of the generalised costs is computed. Based on the number of trips in the status-quo the quantity of increasing or decreasing trips in the variant is calculated. Improvements in the services offered in public transport therefore lead to an increase in traffic demand, as the relative attractiveness of public transport increases compared to road traffic. Consolidation of services on the other hand causes a modal switch from public transport to private cars.

SUPERNOVA’s output:

• Maps show the effects on traffic demand by comparing the number of passengers on each line and the number of cars in each road between status-quo and variante (two variants) geographically.
• Tables show the difference of various indicators in the evaluated variants (service kilometres, service hours, number of passengers, passenger kilometres, car kilometers, etc.). These indicators are available on a very detailed level, every single service trip can be analysed separately.

The economic evaluation is carried out in different systems, with the indicators computed by SUPERNOVA as input.

Projects evaluated

Since its introduction, SUPERNOVA was used to evaluate almost every major project with assumed impact on traffic demand. These ranged from short term projects, such as timetable optimisation, to long term infrastructure development. An overview of some of the most important projects is found below:

• Railjet – the new Austrian premium product in international and domestic long haul traffic
• Vienna Central Station – linking East and West (Bratislava/Budapest-Vienna-Munich/Frankfurt)
• Improvements of the railway system between Vienna and Bratislava 2007/2009
• Vienna S-Bahn-Concept 2007
• Provincial Railway Concepts 2007 (Carinthia, Styria, Upper Austria Tyrol)
• Accessibility analysis comparing the timetables 2003 and 2004

Outlook – increasing the level of detail

From community to counting parish

Although the detail level of the cells in SUPERNOVA is sufficient up to now, there will be a demand to take in more detail in the future. As SUPERNOVA is offered to more and more authorities for evaluating their concepts, the tool requires a higher level of detail, especially in urban areas with complex public transport networks.

The smallest spatial units providing the necessary data are the counting parishes, defined by the Austrian Statistics Department. These smaller cells can be implemented into SUPERNOVA with ease, making it necessary only to redefine the access points to the network (a lot of work).

The main problems will be caused by hardware and software restrictions that were faced during assignment in the current model. As the number of cells would boost from 2,600 to more than 9,000 (multiplier factor 3.5), the possible relations will increase by the factor 12 (from 6.7 million to 81 million).

From NUTS3 to community

A strong focus in the future will be to establish new regional cross border services along with foreign cells having to become more detailed than in the existing model. Furthermore, the traffic flows within those countries must be included into SUPERNOVA, as regional cross border services usually have an impact on inland traffic also. Assuming that some of the neighbouring countries have a lack of data equal to Austrian quality, the main focus will be to collect data of traffic flows between the communities in these countries. A possible strategy will be the cooperation with planning institutions (public or private) already settled in these areas.

Latest activities

Revenue allocation with SUPERNOVA

SUPERNOVA so far has always been used for simulating expected impacts of changes in the transportation system. A completely new question is the allocation of revenues to specific trains, enabling the business units of long haul traffic and regional traffic to divide their revenues. As only a small proportion of tickets are bought with a reservation for a certain train, usually it is only the origin and destination of a trip that is known and not the actual train used by the customer. One method to receive this information is the classic survey, where interviews about origin and destinations as well as the kind of ticket are carried out. This is very cost-intensive and only a certain number of the customers can be interviewed.

As an alternative, the tickets sold can be transformed into trips, containing the origin and destination information. A matrix is built and this is assigned by means of SUPERNOVA, allocating all passengers on an O-D-relation to the possible (combination of) trains on the relation. After a series of tests during Spring 2005, the system proved to be reliable to allocate revenues at least to certain categories of train (EuroCity, InterCity, EuroNight, Regional-Train).

Conclusion

To succeed in the passenger transportation market of the future, a strong know-how in traffic planning will be absolutely necessary. SUPERNOVA helped OBB to boost this knowledge during the past three years and to become the leading public transport company dealing with this topic. The international reputation of SUPERNOVA is visible by looking at the several requests from authorities and other companies to use the simulationmodel for evaluation of their concepts.

Taking these facts into consideration, SUPERNOVA’s importance as an element in the development of new services in public transport is guaranteed in following years.