Model documentation: ASTRA

ASTRA-M (ASsessment of TRAnsport Strategies) is a System Dynamics model suitable for conducting strategic analyses of the transport sector with integrated bundles of policy instruments. The model contains a bottom-up transport module for both passenger and freight transport which is linked with a macroeconomic module, a vehicle fleet module and an environmental module. The macroeconomic part of ASTRA-M includes the development of several economic sectors, such as agriculture, manufacturing industry, supply- and mining industry as well as market and non-market services. The model covers Germany, spatially resolved into NUTS2 regions, corresponding to 39 administrative districts. Annual model calibration periods range from 1995 to 2023 and model projections are simulated from 2024 to 2050. These resulting simulations mainly include the modal split, the future vehicle fleet (both stock and new registrations), and total mileage for both freight and passenger transport as well as total greenhouse gas emissions. To obtain these results, the model proceeds in four consecutive steps: Traffic generation, traffic distribution, choice of transport mode and determination of mileage. From this, an environmental assessment in ASTRA-M is carried out using emission factors per mode of transport. For MIV, vehicles are differentiated according to their size and type of drive (petrol, diesel, electricity, hydrogen, gas). The emission factors are taken from the HBEFA (Manual for Emission Factors for Road Transport). Technological progress is taken into account and a distinction is made between cold start emissions and emissions generated while driving. Furthermore, ASTRA distinguishes between direct (tank-to-wheel) and indirect (well-to-tank) emissions.

Drivers of demand and end-use technologies

Main drivers of transport demand include population development, GDP, energy prices as well as modal choice behavior. While the former three components are exogenously integrated in the model, the latter is endogenously derived and informed by data from both “Deutsches Mobilitätspanel” (BMVI, 2019a) and “Mobilität in Deutschland” (BMVI, 2019b). The modal choice component for passenger transport integrates transport modes like car, bus, train, bike, walking as well as shared forms of mobility such as carsharing, ridepooling and microsharing (shared bicycles and e-scooters). Furthermore, local, short, medium and long travel distances are included as well as urban and rural areas.

Additionally, a technology component covers the differentiation of road vehicle fleets into drivetrain technologies, age classes, vehicle sizes and different emission standard categories. The technologies considered are listed below and cover gasoline, diesel, compressed natural gas (CNG), liquefied natural gas (LNG), liquefied petroleum gas (LPG), battery electric vehicles (BEV), plug-in hybrid electric vehicles (PHEV), fuel cell electric vehicles (FCEV) and trolleys for urban buses and long-distance trucks. The diffusion of these drivetrain technologies is simulated in the agent-based ALADIN model.

Policy instruments

ASTRA-M enables the simulation of a variety of policy instruments for each mode of transport, a detailed overview is presented in Table 1. The implementation of the policy measures starts in different areas of the model. For instance, congestion charge and parking fees are included in the mode choice model as additional monetary costs for car charges. These are incurred per trip. The congestion charge is implemented for the NUTS2 zones of the major cities, while the parking charges apply in all urban regions. The internalization of environmental costs with regard to the CO2 price and the petrol and diesel tax is nominal. The CO2 prices apply to each ton of CO2 emitted, while the tax on diesel and petrol applies per liter. The taxes are added to the price per liter of fuel determined and forecast endogenously in the model and then VAT is added. The subsidies for zero-emission cars are applied in the endogenous fleet model. The subsidies for the various vehicle segments and drive systems are thus taken into account and introduced into the investment decision of households or individuals.

Private carTrucksPublic transportShipsCycling
Commuter allowance Promotion of rail infrastructure expansionTightening of emission limits for pollutantsExpansion of fast cycle connections
Company car taxation Introduction of the “Deutschlandtakt”Obligation to use shore power 
Speed limitsSpeed limitsReduction in track prices  
Fleet targetsFleet targets   
Ban on combustion enginesBan on combustion engines   
Toll exemptionToll exemption   
CO2 adjustment energy tax diesel / petrolCO2 adjustment energy tax diesel / petrol   
Parking management    
Infrastructure expansionInfrastructure expansion   
Table 1: Overview of transport policies per mode of transport

Methods and model framework

For the modeling approach, the logic of the four-stage is applied. The implementation of this approach, which is performed in VENSIM, is graphically represented in Figure 1 for passenger transport and described in the following.

The first stage of traffic generation generates the traffic independently of the change in location caused by it. This is performed by taking into account the development of behaviorally homogeneous population groups (differentiated by age and income groups and NUTS2 zones) and on the basis of trip rates (derived from MID 2017) for four different trip purposes: work, business, private and leisure trips. Additionally induced trips are also added based on macroeconomic developments.

Figure 1: Structure of the modeling of passenger transport module in ASTRA

The second stage uses the traffic generated from the stage before to spatially distribute it. This is done by using origin-destination matrices, which represent the source of the traffic flow and the destination as an interdependency matrix. The spatial division is based on the classification in NUTS-II zones.

The third stage allocates the generated traffic flows to the available modes of transport. With regard to the available modes of transport, ASTRA-M distinguishes between private cars, public transport (bus and rail), bicycles, walking and, with the extension to shared mobility services, also carsharing, ridepooling and micro-sharing (shared bicycles and e-scooters). For this purpose, logit functions are used that calculate the choice decision in an interval from 0 to 1. Information from mobility consumers is used for this purpose. This includes socio-demographic data and data on their mobility tools held such as car ownership. Specific data on the various modes of transport (e.g. travel time and costs) are also included in the calculation. For the passenger model, the mode-choice uses the approach of generalized costs, i.e. the combination of monetary as well as other costs such as value of time. These aspects are integrated into one logit function to endogenously compute respective modal splits.

For the fourth stage, traffic assignment, the traffic flows generated in stage 2 and the modes of transport determined in stage 3 are transferred to mileage required to cover the transport needs to be simulated in the steps before. This enables statements to be made about the utilization of the transport network and where these occur with which means of transport in particular. In ASTRA-M, the first three stages of traffic modeling are run through in detail but due to the fact that the model is no network model, the assignment is carried out via transferring the traffic generated in the first three steps into mileage required. This is due to the focus of the system-dynamic approach on traffic performance and mode choice, which are particularly decisive for the macroeconomic and ecological effects of traffic. To this end, vehicle mileage is calculated in fourth place in the model on the basis of traffic distribution and traffic distribution. By integrating occupancy rates, this enables statements to be made about the vehicle as well as passenger kilometers traveled per mode of transport and trip purpose in the respective spatial resolution.

Similar to passenger transport, the transport demand for freight transport is calculated endogenously in ASTRA-M. Here also a four-step approach is used, which is based on the drivers of freight transport (trade, production of goods, sale to end consumers). The modes of transport include three different types of trucks (light commercial vehicles with <3.5t GVW, trucks with GVW from 3.5t to 12t, and heavy trucks with GVW over 12t including tractor units) as well as rail and inland waterway vessels excluding sea and coastal shipping. The special feature of the ASTRA-M approach is that the four-stage model is directly linked to the fleet models of commercial vehicles, so that policies that change the fleet structure and thus the cost structure of logistics are immediately reflected in transport demand and the modal choice in freight transport. This is not possible, for example, in the previous modeling of the BVWP or the VP2030. Likewise, by linking to policy-sensitive energy price and electricity/energy production models in ASTRA-M, a change in the levy policy for energy sources (e.g. H2, PtL) also has an immediate effect on transport demand.