The challenge for mobility in rural areas is enormous: on the one hand, peripheral areas are under pressure from demographic change, and on the other, a mobility and transport transition must be implemented without undermining accessibility. However, mobility providers and local authorities in these areas are already under strict budgetary control. This is a good reason to look for cost-effective ways of generating transport demand data for future-oriented planning in this context.
In this context, Catchment is supporting the Bad Belzig region in the search for solutions as part of the ‘Residence for MOVERS’ project. The results are presented here in the blog and are also intended to provide practical guidance on how to proceed. The programme was organised by Smart Village e.V. and the ‘residence’ was in the co-working space CocoNat in Klein Glien. It couldn't have been more rural!
Modern traffic models are based on special software, e.g. from PTV Visum or the open source solution MATSim, in order to gain comprehensive insights into traffic flows. In many cases, these models are based on the four-stage model, which is described in more detail below. Extended models, such as activity-based approaches, go beyond the classic scheme by taking into account different groups of people and specific activity pairs - for example, living-working or shopping-living. Multi-agent models further refine these approaches by simulating synthetic populations with detailed daily schedules. These models strive for a system equilibrium that comes as close as possible to real transport demand.
Due to the complexity of our collective mobility behaviour, methods based on this solution remain the best practice for obtaining good traffic and mobility demand data. New data sources and scalable, powerful IT services now enable a different approach, which can be a good first step in estimating transport demand. This development has the potential to make a significant contribution to improving the planning basis in very small municipalities, transport associations and transport companies. This can be seen as a data analytics relaunch of the classic four-stage model.
It comprises the following steps, which we want to carry out at INSPIRE 1km grid level:
Depending on the modelling approach, the sequence of steps can vary or individual steps can be combined.
...can be fun! Estimating transport demand on the basis of population and land use data is common practice in transport modelling. Very specific data bases are often used for this work step. Zensus population data und CORINE-landcover data were used for the Residence Programme.
With the population data, the connection is really figurative: if someone wakes up at home in the morning, they are 80% likely to move away from their place of residence.
Dann werden ein paar Aktivitäten erledigt und es wird der Heimweg angetreten. Was ist aber mit den ganzen anderen Aktivitäten dazwischen? Für eine schnelle Lösung, kann man Schätzwerte für die verschiedenen Landnutzungsformen ansetzten. Zwar ändern diese Werte sich von Fall zu Fall erheblich, jedoch ist das schöne hier, dass zumindest ein europaweit einheitlicher Datensatz genommen werden kann. Insbesondere für kleinere, ländliche Teilräume können dann diese Landnutzungsparameter für die Kalibrierung genutzt werden. Initially, to target the 3.1 trips per person from the MiD 2017 survey in terms of demand generation.
If you want to incorporate very location-specific demand, the integration of OSM POI data, which is more or less standardised nationwide as another open data source, is suitable. For smaller regions, specific volume data can be quickly entered via the POI tables, e.g. for schools, factories or shopping centres. Or fixed values can simply be defined globally for the various POI types. However, the land use parameters should then be corrected again according to the data. PS: If you notice something strange in the OSM data, you can update it yourself.
With the traffic distribution comes the complexity: if you take an area of 100 by 100 kilometres, you end up with 10,000 cells in the INSPIRE grid and thus potentially 100 million source-destination relations. Here it quickly becomes clear that only the bare essentials are included in the calculations. For smaller, rural sub-areas, this can end up being all cells with a demand or above a certain demand threshold. As there is also a lot of ‘nothing’ around Bad Belzig, we have also used all cells with demand here. The demand part of the model is calculated using a simple gravity model that mixes the demand components of the source and destination cells and initially sets them in relation to the airline distance.
If the best routes for the demand from the source-destination relations are now sought, this is known as reallocation. This is where the traffic modelling solutions from PTV or MATsim come into their own, as demand is routed on the transport network in iterations to simulate how demand is distributed as volumes increase. If the waiting time at a node is too long, an alternative route is selected. As we are looking for a simple work-around for pragmatic demand generation, we do not take this route, but use mass calls to routing APIs from Google, Here and Azure, where certain volume effects are already inherent, especially for car traffic.
Hier werden die ersten drei Alternativen der Quell-Ziel-Verbindungen für die verschiedenen Verkehrsträger abgespeichert. Für diesen Schritt wird ggf. nur ein gewichtetes Sample der Quell-Ziel-Relationen genommen z.B. 10%. As part of the project work in the Residence for Movers programme, we have even completely dispensed with demand effects in the route selection for the time being and set up an Open Trip Planner server - also to underline the open source aspect of the programme.
If there is a desire for traffic demand data in the Bad Belzig region, this may be a viable way of obtaining estimated traffic volumes. Further steps that are still open:
Catchment