Comparison with aggregated 4-step model
When comparing ABM Nested Demand to an aggregated 4-step model, the issues considered below are worth highlighting.
- Data requirements
- Skims
- Additional traffic
- Advantages and extended application possibilities of ABM Nested Demand
Data requirements
The data requirements for a 4-step ABM are very similar to those of an aggregated model. For the generation, you need the results of a household survey. This does not necessarily have to have been done in the planning region. A survey in a similar area can be adapted to it (analogous to weighting) using socio-economic data of the planning region.
Model parameters (such as utility functions) can be obtained as in aggregated models by:
- Estimate based on data
- Transfer of existing comparable (also aggregated) models
Structural and population data must be available at the location level. Usually, this is done by disaggregating rougher level data based on building floor plans. The more information is known about the buildings, for example about their use or height, the better the resulting data set.
Calibration and validation are performed in essentially the same manner as for aggregated models. It is not necessary to be more spatially or temporally accurate on a large scale than aggregated models. The main advantage of disaggregation in the 4-step ABM is not so much the small-scale analysis as the more precise skims on the basis of which mode and destination choices are made. And these are often also uncalibrated already significantly better than with aggregated models (which usually use only 24h-skims).
Skims
As with aggregated models, a network model forms the basis for calculating the skims. In contrast to conventional models, more focus is placed on non-motorized traffic, which requires more accurate modeling of the corresponding networks. This includes ensuring pedestrian access to public transportation.
However, this does not mean that every trail must be represented in the model. The reference is still the conventional 4-step model, and a reasonably complete network of bicycles and pedestrian paths should already lead to much more accurate results.
Assignments, which are necessary to derive capacity-dependent skims, are run as usual in aggregated form (e.g. travel time in the loaded network). For capacity-independent skims, no assignments need to be calculated (as a rule, this applies to both non-motorized and public transport). This means that no public transport connections are required for the demand calculation.
Dynamic assignment is explicitly not required to determine dynamic PrT skims. It might even be unsuitable due to its low stability and weak convergence. We recommend using a static assignment for each time range. Compared to the 24-hour assignments on which aggregated models are based, this represents a huge improvement. Since the equilibrium assignments compute extremely fast, the overall computation time of the model is not expected to increase significantly.
Additional traffic
Aggregated models often contain numerous side models such as external traffic, tourist traffic, airport traffic, etc. If such models are not to be modeled disaggregated, this is possible without restrictions. Their impact on disaggregated demand is done, as in aggregated models, by assignment of the common demand of capacity-dependent modes.
Advantages and extended application possibilities of ABM Nested Demand
Of the many advantages and extended application possibilities of ABM Nested Demand compared to aggregated models, only a few are mentioned here in key words:
- A significantly reduced aggregation error
- Consideration of opposing qualities of service in private and public transport
- Better representation of non-motorized traffic
- Morning toll
- Parking space management
- Adding or changing public transport stops without having to correct connectors
- E-mobility: charging in the tour context (range)
- COVID-19: Public transport avoidance only in case of high occupancy rates
- Consideration of mobile working
- Unchanged work location for short-term forecasts
- Individual effect in case of cost changes (value of time)
- Preceded choice of the mobility tool (car ownership, public transport season ticket)
- Analysis of small (not predefined) segments (e.g. single parents)