Professional Documents
Culture Documents
Available online
online at
at www.sciencedirect.com
www.sciencedirect.com
ScienceDirect
Available online at www.sciencedirect.com
Transportation
Transportation Research
Research Procedia
Procedia 00
00 (2017)
(2017) 000–000
000–000
www.elsevier.com/locate/procedia
ScienceDirect www.elsevier.com/locate/procedia
World
World Conference
Conference on
on Transport
Transport Research
Research -- WCTR
WCTR 2016
2016 Shanghai.
Shanghai. 10-15
10-15 July
July 2016
2016
Abstract
Abstract
The
The objective
objective of
of the
the overarching
overarching approach
approach is
is to
to develop
develop aa System
System Dynamics
Dynamics (SD)
(SD) based
based microscopic
microscopic freight
freight transport
transport simulation
simulation
(MFTS) for urban areas. In this contribution, the qualitative SD approach – causal loop diagram – with
(MFTS) for urban areas. In this contribution, the qualitative SD approach – causal loop diagram – with the internal the internal structure
structure of
of
urban freight transport and the stakeholders involved, as well as their interdependencies to each other, are transparently presented.
urban freight transport and the stakeholders involved, as well as their interdependencies to each other, are transparently presented.
On the
On the basis
basis of
of the
the causal
causal loops
loops aa quantitative
quantitative and
and mathematically
mathematically formalised
formalised SD
SD approach
approach will
will be
be derived
derived in
in the
the next
next SD
SD modelling
modelling
steps.
steps. This
This SD
SD model
model enables
enables trend
trend analyses
analyses by
by medium-
medium- and
and long-term
long-term forecasts.
forecasts.
© 2017
© 2017 The
2017 The Authors.
The Authors. Published
Authors. Published
Published byby Elsevier
by Elsevier B.V.
Elsevier B.V.
B.V.
©
Peer-review under responsibility of WORLD CONFERENCE ON TRANSPORT RESEARCH SOCIETY.
Peer-review under
Peer-review under responsibility
responsibility of
of WORLD
WORLD CONFERENCE
CONFERENCE ON ON TRANSPORT
TRANSPORT RESEARCH
RESEARCH SOCIETY.
SOCIETY.
Keywords:
Keywords: Urban
Urban freight
freight transport
transport demand;
demand; Urban
Urban freight
freight demand;
demand; Courier,
Courier, express
express and
and package
package service
service provider;
provider; System
System Dynamics;
Dynamics; Causal
Causal
loop diagramming; Freight transport demand modelling; Stakeholder analysis;
loop diagramming; Freight transport demand modelling; Stakeholder analysis;
*
* Carina
Carina Thaller.
Thaller. Tel.:
Tel.: +49-231-755-8131;
+49-231-755-8131; fax:
fax: +49-231-755-6338.
+49-231-755-6338.
E-mail
E-mail address:
address: thaller@itl.tu-dortmund.de
thaller@itl.tu-dortmund.de
2214-241X
2214-241X ©© 2017
2017 The
The Authors.
Authors. Published
Published by
by Elsevier
Elsevier B.V.
B.V.
Peer-review
Peer-review under
under responsibility
responsibility of
of WORLD
WORLD CONFERENCE
CONFERENCE ON ON TRANSPORT
TRANSPORT RESEARCH
RESEARCH SOCIETY.
SOCIETY.
In the last decades, reurbanisation tendencies of population and economy have been recorded. In 2050, more than
70 percentage of the world’s population will live in agglomerations (OECD 2012). A further trend is the increased e-
commerce activities (EC) of private households, which leads to an increased freight demand of private households.
Another development is that increased storage costs for the urban production industry and the retail sector affect a
decrease of storage capacity in these branches. These tendencies result in a growing share and higher frequencies of
more flexible and small-scaled shipments or deliveries. The increasing shipment frequency leads to growth in the
number of trips of small- and medium-sized vehicles in urban freight transport. Therefore, the total shipment volume
and the total number of trips in urban areas also increase. A higher freight transport volume generates more noise and
greenhouse gas (GHG) emissions and higher utilisation of road infrastructure. Due to higher road infrastructure utili-
zation the infrastructure maintenance costs and the level of traffic congestion are rising. The higher level of traffic
congestion also enhances noise and GHG emissions, risks of traffic accidents and transport lead times in freight
transport. The longer transport lead times affect a higher fuel consumption, which increases the transport costs and
decreases the transport quality (key value for delivery reliability) for the logistics service providers. The lower
transport quality and the higher transport costs reduce the economic efficiency of enterprises. The increased level of
noise and GHG emissions, the number of traffic accidents and the decreased economic efficiency also lead to higher
external costs.
The overarching objective of the following approach is to generate a basic understanding of the system urban
freight transport. The focus is mainly on the interdependencies between, and the decisions and transport processes of,
stakeholders involved at an urban level. For this purpose, transport models are suitable instruments for describing and
explaining the impact of decisions and behaviour of different stakeholders on urban freight transport. Decision-making
processes in transport planning, simulations and assessments of transport planning measures could be carried out by
transport models. (Thaller et al. 2016a) In freight transport modelling, the current models have still a low level of
sensitivity to illustrate behaviour of stakeholders involved and effects of measures due to the large amount of stake-
holders involved, their diversity as well as the existing lack of data (Thaller et al. 2016b). Furthermore, a long-term
forecast for investigating freight transport demand dynamically has not been yet developed in freight transport mod-
elling. In addition, there is no model which considers and explains functional behaviour patterns of, and interdepend-
encies between, the subsystems. These models are often black-boxes which are less traceable. Therefore, a model with
a transparent and detailed structure is necessary to identify these interdependencies or impacts. (Tavasszy et al. 2012)
System Dynamics (SD) enables medium- and long-term forecasts and is a very transparent glass-box model. None-
theless SD does not allow an investigation at infrastructural level. They are established in transport modelling, but
there are only a few approaches which concentrate on freight transport demand. (de Jong 2004, Kuchenbecker 1999,
Weidmann et al. 2012, Thaller et al. 2016a)
For this reason, a SD based microscopic freight transport simulation (MFTS) for urban areas has been developed,
which enables the use of the forecast capability of SD and the detailed resolution of MFTS – in this case Multi-Agent
Transport Simulation (MATSim) developed by Balmer et al. 2009 and extended with an integrated logistics module
by Schröder et al. 2012. The advantages of both methodologies are used to enhance the degree of explanation and the
accuracy of forecasting, as well as to improve the quality and informative value of the impact analysis by assessing
point-in-time and long-term forecasts. (de Jong 2004, Kuchenbecker 1999, Weidmann et al. 2012, Thaller et al. 2016a)
This approach is tested within a case study, which focuses on the interdependencies of the stakeholders’ courier,
express and package service providers (CEP service providers), private households as end consumers and transport
users as well as the retail sector at an urban level.
In this contribution, the SD model developed will be explained and transparently illustrated by causal loop dia-
grams, which will show the internal structures of the submodules and the feedbacks between them in a qualitative
way. The SD model is able to give qualitatively-valid statements about developing tendencies and dynamics as the
effects of freight transport development, economic activities and population development, as well as environmental
effects at urban level. The primary focus is on the decision processes and behaviour of the freight demand (e.g. private
households, retailers, shippers) and the resulting freight transport demand of the logistics service provider (LSP; in
this case CEP service provider). The freight transport demand also effects on the overarching traffic situation and
environment at an urban level. The investigation area is the German city Berlin. (Thaller et al. 2016a)
Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094 1077
Thaller et al. / Transportation Research Procedia 00 (2017) 000–000 3
Following the problem definition and the derived research objectives, the current state of the art in SD modelling
will be shown in chapter 2. In chapter 3, the methodological approach and SD modelling procedure will be discussed
in detail. In chapter 4, we will present the qualitative SD model developed (causal loop), which describes the urban
freight transport landscape with the relevant stakeholders and their interdependencies. In the summary, the contribu-
tion will be concluded in chapter 5 by a short conclusion and outlook for further research steps in this project.
In this chapter, the state of the art in SD modelling in the research fields of transport planning, spatial development,
economics and social sciences is shown. The approaches presented are investigated in detail and consideration is given
regarding if they are suitable for our own research focus.
The methodology SD was developed by Jay W. Forrester (1961). The foundation for his Industrial Dynamics model
is the General System Theory and Control Theory, in which he investigates economics, business and organisation
systems. The methodological approach serves as a basis of explanation to illustrate effects of decisions in complex,
dynamic systems. In particular, the time functions of the SD approach are emphasised. In the focus of interest are the
internal structures, the system behaviour and the external interactions or feedbacks between different stakeholder
systems. (Rosenberg et al. 2012, Bossel 2004a,b, Sterman 2001, Weidmann et al. 2012) The specific feature of SD is
its non-linear feedback structures and functions. For this reason, the interdependencies between system submodules
have to be identified and illustrated in an iterative modelling procedure. (Thaller et al. 2016a, Weidmann et al. 2012)
Freight transport at an urban level has not been sufficiently taken into account in SD modelling. In the following,
the current state of the art in the research fields (mentioned above) illustrates this recognition. Most of the transport
planning approaches for urban or regional areas only consider motorised individual transport (MIT) and public
transport. They examine the internal structures of these systems at an urban level and their dependencies on economics
and spatial development. Besides this they investigate the impacts of transport policy measures on transport demand
and spatial development. However, freight transport as an essential part of the total urban system is not considered in
these approaches. (see Lee 1997, Heimgartner 2001, Raux 2003, Raux et al. 2007, Emberger 1999, Pfaffenbichler
2003, Mousseau et al. 2012, Wang et al. 2008, Zhan et al. 2012, Ruutu et al. 2013, Hong et al. 2011, Han and Hayashi
2008)
Further SD models focus freight transport either very basically as a part of the holistic system or more explicitly at
a broader regional, national or international level (see Schade et al. 2000, IWW 2000, Schade 2004, Kuchenbecker
1999, Fiorello et al. 2006). Further approaches analyse the complete freight transport market or the interdependencies
between logistics strategies and freight transport demand along the entire supply chain (see Gacogne 2003, Salini and
Karsky 2002, Weidmann et al. 2012, Aschauer 2013, Aschauer et al. 2015, Kühn and Krail 2013, Meyer Sanches et
al. 2013).
The decisions of and interdependencies between customers, retailers and suppliers are already analysed by SD
approaches from economic research perspective (e.g. Villa et al. 2013). However, these approaches do not investigate
logistics and freight transport. In regard to socio-demographic and socio-economic development of population and
private household structures, there are many approaches which examine society and its characteristics (e.g. mobility
and also consumption behaviour of different types of households) (see Bossel 2004a,b,c, Engel 2010, Jin et al. 2009).
The discussed approaches in SD modelling are examined within an extensive literature analysis to develop a new
SD approach. They are adapted to the own research issues and – if necessary – are extended.
In the following chapter, the theoretical foundation of the methodology SD is explained. Furthermore, the SD
modelling procedure is described and subsequently our own methodological procedure is derived from the consider-
ation of the research focus.
1078 Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094
4 Thaller et al. / Transportation Research Procedia 00 (2017) 000–000
To meet all the requirements for SD, we need instruments for modelling. These consist of the system elements as
well as the feedback loops. A dynamic system has to handle a rich complexity of elements. Bossel (2004a) establishes
the relevant principle that the system’s behaviour can only be illustrated if the characteristics of the system elements
as well as their interdependencies or relations are correctly reproduced. (Bossel 2004a, Kapmeier 1999, Sterman 2001)
The categories of system elements are constants, stocks, flows and intermediates or auxiliaries. Constants are system
parameters, which are provided at the outset and for this reason cannot be changed. Stocks are also designated as
storage parameters and they are constantly changed by the in- and outflows or flows in the simulation process. They
are formalised in integrals. For this reason, they can only be investigated in relation to time. Furthermore, the flow
variables play an important role since they determine and influence the change of the stocks in each time interval.
These flows can be formalized in differential equations. Intermediates or auxiliaries are formalised in algebraic or
logical functions. (Bossel 2004a, Weidmann et al. 2012) The dynamics in the modelling results from the so-called
feedback loops. In this case, it is differentiated between positive and negative loops. Positive loops reinforce the events
in the system and lead to growth. From mathematical perspective, the sign is maintained with regard to positive feed-
backs, from which a destabilisation of the system can often result. In contrast, the effect of the negative loops is
inversely proportional and leads to a change. These negative inhibitory loops are often self-regulating and limiting.
Consequently, they exert a stabilising effect on the system. Considering the mathematical perspective, a change of
sign is taking place. Although, there are only two types of feedback loops, a complex system can include more than
thousand loops. (Bossel 2004a, Sterman 2001) A subdivision of the total system in independent submodules should
therefore be performed to illustrate the complexity of dynamic systems in a more transparent manner. This is an
independent part of the modelling procedure. The advantage of these submodules is that they can be tested inde-
pendently from the total system. Only after checking the submodules, they are again linked to a total system. (Bossel
2004a) The objective of the SD modelling may be to determine a policy. They are decision rules which are applied in
the daily routine. They are informally determined by societal and peer pressure or in a written document. For the initial
modelling procedure it is important to realistically illustrate the currently prevailing policies. (Kapmeier 1999)
In the following chapter the procedure for developing our SD is discussed in comparison to the existing modelling
procedure approaches of Forrester (1961) and Kapmeier (1999). Forrester (1961) originally developed a 6-step ap-
proach, which has been extended by Kapmeier in 1999 to a further intermediate step (Kapmeier 1999). We have now
adapted and extended the previous procedure to a 10-step one.
First, the research problem and the specific research objectives – with the theoretical hypotheses – are identified
and determined (see chapter 1). In the second modelling step, the examined system and its behaviour – in this context
the urban freight transport system with its interdependencies to urban freight demand and freight transport demand –
is verbally described on the basis of a literature analysis. Against this background, existing SD approaches are ana-
lysed, adapted and extended for the following SD approach. (Thaller et al. 2016a, Kapmeier 1999) The third modelling
step considers the development of causal loops. This step can also be termed qualitative modelling or causal loop
diagramming. In this step the interdependencies of all system elements are modelled, albeit without defining mathe-
matical formulas. Although the causal loop diagrams are not provided by Forrester (1961), Kapmeier (1999) intro-
duces them as an intermediate step. His justification for this is that the development of causal loop diagrams facilitates
the SD user to familiarise themselves with the matter and to improve their instinct for cause-effect relationships. In
this research approach, they are used to describe and explain the internal structure of urban freight transport and its
subsystems (e.g. shipper, logistics service provider, retailer, private households). In this step, the interdependencies
between the stakeholders described are also modelled. The fourth step is integrated as an intermediate step by us to
validate the causal loop diagrams by experts from the field of transport planning, transport logistics, economics, social
sciences and SD modelling. In the fifth step the validated causal loops are converted into a stock-flow chart - the
quantitative SD approach. In this step the mathematical functions are formulated to quantitatively define the internal
structure, the behaviour, the decisions and the interdependencies between the submodules. The SD approach is defined
by differential equation systems and their internal elements stocks, flows and auxiliaries. (see chapter 3.1) This stock-
Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094 1079
Thaller et al. / Transportation Research Procedia 00 (2017) 000–000 5
flow chart may also be used for simulations. (Kapmeier 1999, Weidmann et al. 2012) In step 6, the mathematical
formalised SD model has to be parametrised with the historical data of a selected investigation area, in this case a
German major city (Berlin). We have implemented this step, because the effort of collecting secondary data is very
time-consuming and should therefore not be underestimated. In step 7, first simulations of the reference scenario may
be performed. Within this modeling step, it is necessary to validate the system by structure and behaviour tests. The
simulation should illustrate the potential difficulties at the end, which may have arisen during the implementation of
changes in the real system. For this reason, the initial verbal formulations and the mathematical assumptions have to
be subsequently retrospectively checked. At this stage, although the model is ready for simulation, it is perhaps not
sufficiently behaviour sensitive enough. At this point, it is necessary to prove the model accuracy of the reference
scenario by a sensitivity analyses (step 8). Following the sensitivity analyses to enhance the behaviour sensitivity of
the system, alternative policies and structures (developing drivers) are implemented in future scenarios (step 9). These
are determined by empirical values or by an iterative procedure. In this instance, alternative policies could be transport
policy or urban logistics measures. They could be implemented in future scenarios. Developing drivers of population,
economics and logistics could also be considered. After implementing these measures or developing drivers, the sce-
narios could be simulated. The resulting outcome of these simulations can be compared with the results of the refer-
ence scenario and correspondingly assessed. (Thaller et al. 2016a) In step 10, the discussion and debate, the simulation
participants work on the consistency in views. By further interviews or workshops with experts the final models and
the results are validated. This step is the most difficult stage in the project since a high degree of unanimity for simu-
lation results has to be achieved. (Kapmeier 1999) In the 11th step, the changes in the real system are implemented.
The modeller is here only able to control or monitor if the modelling is useful and without errors. Therefore, the long
time period of changes has to be potentially considered. The verification, if the final changes are induced internally or
externally, is thus difficult to assess. (Kapmeier 1999) We have excluded this modelling step for our SD approach due
to time restrictions and fewer possibilities to implement transport policy and urban logistics measures as a research
institution. Our SD model serves as a laboratory environment to investigate the potential development of the system
and the effects on freight demand and freight transport demand. In addition, it is very important to also perform
iterative improvements and changes of the system by feedbacks at every stage (Kapmeier 1999). The current state of
this project is that the first six modelling steps are finished and the 7th step is currently in progress. In this step, the
validation (by behaviour and structure tests) and the calibration (based on historical data) are performed. The last three
modelling steps, 8-10, will be conducted and finalised within the following year.
In the next chapter, the internal structure of the submodules and their interdependencies are qualitatively described
and illustrated by causal loop diagrams.
4. System Dynamics approach for describing and explaining the urban freight demand and freight transport
demand
In the following chapter, the developed SD model is first qualitatively described in detail. On this basis, the quali-
tative SD model, transparently illustrated by causal loop diagrams, is derived and explained. The causal loops show
the internal structures of the submodules as well as the feedbacks between each other. The total SD model is therefore
subdivided into as small submodules as possible to guarantee the transparency and traceability for the reader. The
interdependencies between the parameters observed are indicated with link polarities (+/-). A feedback with a plus-
sign has a positive reinforced effect on the dependent parameter. A minus-sign, in contrast, specifies the interdepend-
ency between two parameters as negative inhibitory. Observed parameters within the causal loop diagramming are
given in italics in the following descriptions.
In this case study, we focus on the following stakeholders.
• CEP service provider as a representative for a logistics service provider (LSP)
• Private households (PH) as end consumers and transport users
• Retail sector (R) as a representative for the urban economy (UE)
An assumption is that the shipper (e.g. wholesaler [WS], production industry [PI]) has an infinite capacity to pro-
vide the freight demand of private households and retailers. For this reason, the internal structure of the shipper is not
considered in this approach.
1080 Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094
6 Thaller et al. / Transportation Research Procedia 00 (2017) 000–000
4.1. Socio-demographic and socio-economic structure, household structure, consumption and mobility behaviour
In the following subsection, the qualitative description of the submodule population is performed. The population
can be differentiated by various forms (e.g. age cohorts; household structure, socio-demographic and socio-economic
characteristics, consumption and mobility behaviour patterns as well as way of life [sinus milieus]). In this approach,
we differentiate the population behaviour and the internal structure of population in socio-demographic and socio-
economic structure, household structure as well as the consumption and mobility behaviour. The following submod-
ules of the population are based on the research approaches of Lee (1997), Heimgartner (2001), Raux (2003), Raux et
al. (2007), Emberger (1999), Pfaffenbichler (2003), Mousseau et al. (2012), Wang et al. (2008), Zhan et al. (2012),
Ruutu et al. (2013), Han and Hayashi (2008), Schade et al. (2000), IWW (2000), Schade (2004), Kuchenbecker (1999),
Fiorello et al. (2006), Villa et al. (2013), Bossel (2004a,b,c), Engel (2010) and Jin et al. (2009). These approaches
have to be adapted and extended for our specific SD modelling approach.
The population develops positive reinforcing by a positive birth and mortality balance on the one hand and by a
positive migration balance on the other. With the number of population and the number of cars per person we can
derive the number of cars. The higher the total number of cars, the higher the number of car trips or passenger transport
volume will be. The positive reinforcing feedback leads to higher passenger kilometres performed. The relationship
is, in this case, positive reinforcing. The passenger kilometres performed are influenced positive reinforcing by the
average distance travelled by car. Passenger kilometres performed effects on GHG emissions and the fuel consump-
tion of MIT. GHG emissions are also dependent on fuel consumption of MIT. Both parameters are positive reinforced
by passenger kilometres performed and the level of traffic congestion. These assumptions conclude the socio-demo-
graphic characteristics of the population and their mobility behaviour. The population development also influences
the number of households. In the quantified SD approach, the population will be differentiated in three different house-
hold cohorts according to household size (e.g. number of households with one person, with two persons and with three
and more persons).
The positive employment rate is an indicator of the economic prosperity of a city. This means that the economy
guarantees a positive employment development. The employment effects on unemployment in a negative inhibitory
feedback. The higher the employment rate, and also the level of wages, the greater is the migration balance will be. In
contrast, a high unemployment rate affects a migration balance in a negative inhibitory way. In addition, the positive
employment rate leads to a positive income development of private households and a positive level of wages develop-
ment. These parameters and their interdependencies describe the socio-economic characteristics of the population.
The income development effects positive reinforcing on the parameters disposable income for consumption, re-
quirements on quality and freight demand by EC activities. The disposable income for consumption influences positive
reinforcing the freight demand of the private households in total. In addition, the requirements on quality also affect
the freight demand. The total freight demand of private households is differentiated in freight demand in location-
bound retail branch R and freight demand by EC activities via the internet to the PI or WS (shipper). In regard to
freight demand we will determine which goods are purchased. These goods will be differentiated in commodity groups
by the Nomenclature uniforme des marchandises pour les statistiques de transport 2007 (NST-2007, Statistisches
Bundesamt 2008). Furthermore, the commodity groups can be divided into short, medium and long-term needs. This
subdivision is necessary in order to answer the question of which goods are purchased from the traditional R or which
ones are ordered by an online purchase.
In this instance, we have to ask which commodity groups are suitable for the location bound R purchase and which
ones for EC. The higher freight demand of private households from R, the more the share of additional trips for private
consumption will be. Therefore, the number of car trips (transport volume of MIT) increases as a result of these
additional trips. Both parameters freight demand R and freight demand EC are negative inhibitory to each other. The
household size affects both parameters freight demand R (positive reinforcing) and freight demand EC (negative in-
hibitory). In relation to the EC the client has the choice between different shippers or suppliers (home delivery, indirect
delivery via WS, online R).
In respect to EC purchases, the client has the choice between different CEP service providers (who will deliver the
package? e.g. DHL, Hermes, UPS), between different delivery services (e.g. home delivery, service point, package
station/box, depot, etc.) and the delivery time windows (when will the delivery be? e.g. same day delivery, express,
standard). These possibilities of choices should be also illustrated in the quantitative model by each share of the freight
Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094 1081
Thaller et al. / Transportation Research Procedia 00 (2017) 000–000 7
demand. The increased EC freight demand increases the order amount per order frequency – by the hypothesis – the
higher the order frequency the lower the order amount per order frequency. This relationship is a negative inhibitory
feedback loop. By the higher order amounts returns of the specific groups of commodities (e.g. clothes) are more.
This relationship is, in this case, a positive reinforcing feedback. Returns are also conducted by the higher require-
ments on quality. The small sized deliveries have a positive reinforcing effect on the delivery amount and frequency.
The higher the delivery amount, the higher the number of trips of LSP will be. In addition to the higher deliveries
higher return rates are observed in the course of EC activities. The higher freight demand EC of the customers de-
creases the freight demand in location-bound retail and therefore the trips for purchasing. In addition, additional trips
to pick up locations (package stations, post office branches) for the packages are generated by EC demand. These
assumptions explain the consumption behaviour of private households based on their socio-economic characteristics
(e.g. income). To sum up, the freight demand of private households generates additional car trips to R for private
consumption. Through EC activities or online orders to WS and PI freight transport volume of the LSP is generated.
In Fig. 1 the causal loop diagram describes the socio-demographic and socio-economic structures of the population
as well as their mobility and their consumption behaviour being presented. The causal loop shows the
interdependencies between the different parameters observed. In this causal loop parameters in grey and in a content
bracket are shadow variables, which are already existing parameters of other submodules.
The mathematical formalised SD model is derived from the assumptions of the qualitative causal loop diagramming
(described above). The parameters marked in red in the causal loop diagram are excluded for the quantitative SD
approach to define the boundary of the system and to limit the research focus for the later simulations.
-
Unemployment
Employment - +
-
+
+ +
+
+
Requirements
Number of cars Disposable income on quality
for consumption Delivery
+ reliability
+ + +
Freight +
demand Freight demand +
Number of car trips e-commerce
Fuel + -
+ + - Order amount
consumption car + +
+ private e-commerce
Passenger kilometres - + Order frequency
Motorised Individual + Additional trips for -
+ private consumption + - private e-commerce
Transport
+ + Freight demand Returns
<Level of traffic retailer Household
+ Trips to pickup - + size
congestion> locations (packstation,
post office)
+ GHG emissions + <Freight demand
Motorised Individual e-commerce>
Transport
Fig. 1. Causal loop of socio-demographic and socio-economic structure, household structure, consumption and mobility behaviour of population
(Source: own diagram based on Lee 1997, Heimgartner 2001, Raux 2003, Raux et al. 2007, Emberger 1999, Pfaffenbichler 2003, Mousseau et al.
2012, Wang et al. 2008, Zhan et al. 2012, Ruutu et al. 2013, Han and Hayashi 2008, Schade et al. 2000, IWW 2000, Schade 2004, Kuchenbecker
1999, Fiorello et al. 2006, Villa et al. 2013, Bossel 2004a,b,c, Engel 2010, Jin et al. 2009)
For this reason, we have excluded the parameters employment, unemployment, level of wages and fuel consumption
of cars, because these values are not in the interests of the research focus. Furthermore, the requirements on quality
have not been considered in the quantitative approach, since it is a qualitative characteristic, which cannot be quanti-
fied. In addition, the delivery reliability have not been implemented in the quantified approach, since the private
customers of the CEP service providers could not be integrated as sinks in the system and thus the delivery reliability,
respectively the punctuality of deliveries, could not be illustrated by SD. Moreover, in this approach we only focus on
1082 Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094
8 Thaller et al. / Transportation Research Procedia 00 (2017) 000–000
MIT and do not concentrate on public and non-motorised transport. In the quantified SD approach, we have differen-
tiated according to three household sizes. In regard to household structure, we have not considered the structure of
specific sinus milieus and their specific consumption behaviour. Sinus milieus are mostly described in the state of the
art in a very qualitative way due to the limited data for its description. In respect to freight demand we have not
subdivided in different commodity groups (NST-2007, Statistisches Bundesamt 2008), since the insufficient data
quality have not enabled this implementation. Furthermore, existing freight demand, defined as fiscal values, would
have to be transformed by value-density-transformation to volume units of the goods demanded. Afterwards, the vol-
ume units would have to be transformed to package units. In this case study with focus on CEP service provider this
procedure would be extremely difficult to implement in practice due to the lack of data. To sum up, the households
have differentiated according their household sizes.
In the next subsection, the internal structure of the freight demand behaviour of R is qualitatively discussed on the
basis of a causal loop diagram. This causal loop diagram is derived from Villa et al. (2013) and Bossel (2004b,c) as
well as Schade et al. (2000), IWW (2000), Schade (2004), Kuchenbecker (1999), Kaczmarek et al. (2004) and Fiorello
et al. (2006).
The freight demand of private households negatively inhibitory influences the storage stock of R. This results in a
negative inhibitory effect on the supply gap or storage deficit of R. First, R controls his storage stock. If he determines
a supply gap, he carries out a follow-up order. In this context, the lower his storage stock or his storage capacity, the
higher his order frequency will be. The relationship between storage stock and order frequency is a negative feedback
loop. By the increased freight demand of PH and the resulting increase of the storage deficit the total orders per order
frequency of R to the shipper (WS or PI) are consequently increased. The more orders per order frequency, the lower
the order frequency will be. This relationship is a negative inhibitory loop. The increased orders have a positive rein-
forcing effect on the backlog of the shipper (WS or PI). The WS’s backlog (shipper) is also increased by the orders of
UE and PH. In this case, the WS also controls his storage stock based on his backlog. If there is enough stock, the
shipments can be delivered. The shipment amount to be delivered is correspondingly commissioned and delivered to
the client. If there are not enough storage stocks, a storage deficit or supply gap is determined, which results in orders
to PI and UE. The higher the backlog of WS, the greater the accumulated deliveries to the clients (in this case to R)
will be. The higher the shipment to be delivered, the lower the shipper’s capacity (WS) will be. First, a target delivery
time (determination of delivery date/ time window) is agreed between the client R and the shipper WS. Due to supply
gaps by the shipper WS delivery delays can result. The higher the shipments to be delivered from WS to R/received
deliveries R from WS, the greater the number of transports from WS to R will be. Capacity deficits or supply con-
straints by the shipper (for example of WS) lead to the necessity to either build up or extend the storage stock or
capacity, or to compensate the storage deficit or supply gap by increasing the order frequency to PI or UE. The exten-
sion of the capacity increases the time delay until the extension is finalised. More orders result in a higher delivery
frequency. The submodules PI and UE are structured on the same principle due to the submodules being interdepend-
ent on each other. They represent the supply side as well as the demand side.
In this case study, the freight demand behaviour of R has been only considered with regard to the total goods
demand per time interval for the mathematical formalisation of the quantitative SD model. The implementation has
been difficult due to the existing lack of data on the one hand and the complicated transformation procedure (of fiscal
values to volume units to package units) on the other hand. Although the interdependencies described above could be
applied for a business perspective to illustrate the structure of a single retailer, but the specific variables of an entire
sector – in this case retail sector – could not be parametrised due to the lack of data (e.g. storage stock of R, storage
capacity). Since the retailers as customers of the shippers could not be integrated as sinks in the system and thus the
parameters to derive the delivery reliability (e.g. actual delivery time, delivery delay, target delivery delay) could not
be realistically calculated by SD. It is not necessary for the SD quantification to integrate the shipper’s capacity, his
supply gap and the respectively derived time to build up the capacity and the resulting delivery delay due to the infinite
capacities of the shipper. If the LSP has a delivery delay and does not renege the target delivery delay, the delivery
reliability is decreased. Due to delays he has to pay penalties to the shipper. The only parameter considered regarding
shipper is the shipper’ backlog, which is a sum of freight demand EC of private households and total orders of R.
Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094 1083
Thaller et al. / Transportation Research Procedia 00 (2017) 000–000 9
To sum up, only order frequency of R and the total orders per order frequency of R to the shipper as retailer’s
freight demand and the shipper’s backlog are used for the mathematical formalisation. R’s freight demand behaviour
as well as his relation to private customers and the shipper is shown by a causal loop diagram (see Fig. 2).
4.3. Freight transport demand behaviour and supply development of the logistics service provider (LSP)
In the following the internal structure of LSP, in this case the CEP service provider, is described qualitatively and
in-depth by causal loop diagrams. The system of this stakeholder is subdivided in the following submodules.
• Freight transport demand behaviour and freight transport capacity development
• Shipment amount and freight transport volume development
• Freight transport stock development
• Tour characteristics and transport lead time development
• Road mileage performed for last mile development
• Freight transport cost development
First, a brief introduction in the general structures of freight transport service providers is carried out.
The distribution from goods outward, transport, transshipment, storing, commissioning to the incoming goods from
the source (shipper) to the customer is either handled by an internal transport with its own fleet park or by a logistics
service provider (e.g. freight forwarder, transport service provider, CEP service provider), who undertakes these tasks
as a service. An enquiry for a transport service from the shipper to the service provider is an order to deliver shipments
from the source to the sink (to the customer). The shipper awards a delivery contract to the logistics service provider.
Fig. 2. Causal loop of retailer’s freight demand behaviour (Source: own diagram based on Villa et al. 2013, Bossel 2004b,c, Schade et al. 2000,
determines the lot size choice by the transport capacities. Within the transport chain split the LSP plans how the
transport chain is split. Therefore, it is necessary to have an overview of transport network hubs and distribution
centres. (Thaller et al. 2016b) In this context, we have to consider if the transport from source to sink is carried out by
internal transports via road (directly without a transshipment), or if the transport chain is split into first run, line haul
and last mile, and the goods are transshipped at logistics hubs. First run and last mile distribution have a spatial radius
for the service territory of 100 to max. 150 km. (Weidmann et al. 2012) If the transport chain is split, different transport
modes (among trucks) can be used for the line haul (e.g. air, rail and waterway transport). In the framework of tour
planning and route choice the different transport networks and distribution hubs as well as the sinks (addresses of the
clients) have to be determined. (Thaller et al. 2016b) In this approach, the focus is on the road freight transport in
urban areas. The first run and line haul are, however, considered by a share of transport costs for the first run and line
haul. The line hauls via rail, air and waterway are covered by external inputs (constants) in the system city. The
characterisation and internal structure of the LSP are described in the following subsections.
4.3.1. Freight transport demand behaviour and freight transport capacity development
In this subsection, the submodule freight transport demand behaviour and freight transport capacity development
are described in a qualitative fashion. This submodule concentrates on the freight transport demand (given in volume,
mass or weight of different goods), to be transported, depending on a finite freight transport capacity of LSP. The
following causal loop diagram illustrates the freight transport demand and the freight transport capacity development
(see Fig. 3).
The shipments to be delivered from the shipper (PI, WS and UE) to the customers R and PH are the transport
demand or orders of the shipper to the logistics service provider. He awards transport contracts to LSP. The higher
the transport demand of the shipper (backlog of the shipper), the greater the backlog of LSP, in this case the number
of contracts, will be.
+ Transport
Backlog Logistics demand/orders
+ (shipper)
Service Provider
(number of contracts) Desired transport
+
services
- +
<Truck load - <Shipper's
+ -
capacity> - backlog>
- -
- + <Target delivery
delay>
Transport demand +
<Accumulated to be executed Transport
transport lead time> - capacity
- + Delivery Delay
+ +
LSP
Actual delivery <Transport
Expected transport time
lead time + stock>
+
Fig. 3. Causal loop of freight transport demand behaviour and freight transport capacity development (Source: own diagram based on Villa et al.
2013, Weidmann et al. 2012, Kaczmarek et al. 2004, Kuchenbecker 1999)
This leads to a high rate of desired transport services for LSP, which influences the transport capacity negative
inhibitory. The transport capacity is determined by the product of the number of trucks (LSP’s truck fleet or transport
stock) and the truck load capacity. Therefore, the maximum possible transport capacity which can be transported is
defined. This feedback between both parameters is negative inhibitory. Therefore, LSP controls his transport capacity
depending on whether he can provide and execute the transport service. In principle, the higher the backlog of LSP,
the lower the transport capacity of LSP will be. If the contract can be performed, the contract is accepted and executed.
Through the executed transport services the transport capacity again decreases. Between the shipper and the logistics
service provider an appropriate target delivery delay is agreed. In this case, the target delivery delay also effects on
the desired transport service, transport capacity and the transport demand or orders of the shipper.
Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094 1085
Thaller et al. / Transportation Research Procedia 00 (2017) 000–000 11
The higher the transport contracts or orders and the lower the transport capacity, the sooner the actual delivery
delay will be. In cases of deviation from the target delivery delay, the shipper will award less transport contracts or
orders to LSP and will choose another LSP the next time. In addition, LSP has to pay penalties to the shipper for the
delays or late arrivals of the deliveries. Transport capacity is also dependent on the existing transport stock and the
truck load capacity of each vehicle of LSP in a positive reinforcing manner. The higher the transport capacity, the
greater the transports to be executed will be. The delivery delay is dependent on the desired transport services of the
shipper in a positive manner as well as on the transport capacity in a negative inhibitory one. Moreover, the target
delivery delay influences the delivery delay. The delivery delay increases the actual delivery time, which is also pos-
itive reinforced by and dependent on the expected transport lead time and the accumulated transport lead times. The
actual delivery time decreases the transports to be executed. (see Villa et al. 2013, Weidmann et al. 2012, Kaczmarek
et al. 2004, Kuchenbecker 1999)
For the quantified SD approach, we have also waived the criteria, which describe the delivery reliability (e.g. target
delivery delay, delivery delay, actual delivery time, expected transport lead time) for the same reasons explained in
chapter 4.2.
provided the same day. Therefore, returns do not usually induce higher transport costs. Rather, they have an effect on
higher logistics costs for the examination of goods, repackaging and commissioning at the logistics hubs. The causal
loop illustrates the submodule shipment amount and freight transport volume development (see Fig. 4).
For the quantification of this submodule we have excluded the pressure to consolidate, since it is a qualitative
value, which is based on an imprecise estimation. Returns have not been integrated, since they have no effect on the
development of road freight transport volume. In addition, the order frequency R and EC have been already consid-
ered for the quantified approach in other submodules (see chapters 4.1 and 4.2).
+
Spatial dispersion
of clients + Total number of
Transport deliveries Average distance
Truck costs between stops (density
per hour distance per tour
+ + of stops)
Transport lead
time costs Number of +
<Average speed
+ deliveries per tour
on road> - - <Spatial dispersion
+ - <Road freight - of clients>
<Road freight +
+ - transport volume>
transport volume> Accumulated
transport lead time Number of stops
per tour
+ Transport lead -
+ Drop factor
time per tour
+ Share of first delivered
- package units to end
Average stop consumer
time
+
Time intensity of
delivery service/
transport concept
Fig. 6. Causal loop of tour characteristics and transport lead time development (Source: own diagram based on Moder 2010, Aschauer 2013,
Aschauer et al. 2015, Weidmann et al. 2012, VDI 2002)
The causal loop of the submodule tour characteristics and transport lead time development is shown in Fig. 6.
In the quantitative SD approach the transport lead time costs and the truck costs per hour have not been considered,
because they are already indirectly included in the personnel costs (fixed costs) and in the variable costs (dependent
on road mileage performed for last mile) (see Fig. 8). Furthermore, the spatial dispersion of clients, the share of first
delivered package units to end consumers and the time intensity of delivery service (transport concept) have not been
integrated in the quantification, since there is no specific data available regarding these parameters. The spatial dis-
persion of clients is also difficult to measure (spatial relations cannot be illustrated or calculated by SD approaches).
The feedback from average speed on road to accumulated transport lead time has also not been included in this
approach, since it has already been considered in the relation between average speed on road and transport lead time
per tour. In addition, we have waived the impacts of road freight transport volume and total number of deliveries on
the number of deliveries per tour, since these feedback loops to the number of deliveries per tour destabilise the total
system.
and Meyer Sanches et al. (2013), and are only broadly valid for the line haul. In relation to the quantified last-mile
model, the modal split has been excluded.
For this approach, only the motorised freight transports on road for the last mile distribution are considered. Road
mileage performed for last mile is calculated by the product of road freight transport volume and transport distance
per tour. The total road mileage performed is positively influenced by the road mileage performed for last mile and
the average operating life of vehicles. Transport performance is positive reinforced by the road mileage performed
for last mile and the shipment amount. The total amount of fuel consumed is derived from the fuel consumption per
truck and kilometre and the road mileage performed for last mile. Furthermore, the total amount of consumed fuel is
also dependent on mass or weight of the load. The total amount of fuel consumed is positive reinforced by the level of
traffic congestion. The GHG emissions are positive reinforced by the road mileage performed for last mile, the level
of traffic congestion, the total amount of fuel consumed and the GHG emissions per truck and kilometres. Aschauer
(2013) and, moreover, Aschauer et al. (2015) consider the resulting costs for GHG emissions, which are derived from
the polluted GHG emissions in road freight transport and the internalisation of emission tax. We have excluded these
costs for the quantified SD approach. The reason for this decision is provided in chapter 4.3.6.
Fig. 7 illustrates the causal loop, which describes the development of road mileage performed for last mile.
For the quantification of the SD approach the parameters modal split, average transport distance in total, mileage
performed in line haul and mileage performed in total as well as the consolidation pressure (see Aschauer 2013,
Aschauer et al. 2015) have been excluded. We have also ignored the impact of level of traffic congestion on total
amount of fuel consumed and on GHG emissions of road freight transport.
For further modelling steps, we have to consider the modal split with regard to road freight transport. Thus, different
vehicle types with different truck load capacities and forms of propulsion (e.g. e-mobility) as well as non-motorised
vehicles (e.g. cargo bike) can be used for the last mile distribution.
(catalogue price-deduction) and transport stock have a positive reinforced effect on unit-of-production depreciation.
In this case, purchase price (catalogue price-deduction) is derived from the product of purchase price for one truck
and number of new purchased vehicles. Maintenance and repair costs are positive reinforced by the total road mileage
performed and average operating life time of the vehicles. If a municipality implements a city toll, toll costs are
influenced and positive reinforced by the toll charge and the road mileage performed for last mile. The fuel consump-
tion and fuel price development positively reinforce the fuel costs. Lubrication costs are dependent on and positively
reinforced by road mileage performed for last mile and fuel costs. Variable costs are calculated by the sum of fuel
costs, lubrication costs, tyre costs, unit-of-production depreciation, maintenance and repair costs and toll costs.
Time depreciation is positively reinforced by the transport stock and the purchase price. The parameters residual
value, average operating life of vehicles and tyre prices effect negative inhibitory on time depreciation. Interest costs
are positive reinforced by purchase price, residual value and transport stock. Taxes and insurances are dependent on
transport stock. Personnel costs are dependent on the transport stock and the accumulated transport lead time. Fixed
costs are calculated by the sum of time depreciation, interest costs, taxes and insurances, personnel costs as well as
other costs. Total transport costs are resulting from the sum of fixed and variable costs. (Wittenbrink 2011) In Fig. 8
the causal loop illustrating the development of freight transport costs is shown.
We have excluded for the quantified SD approach the parameter share of transport costs for line haul. Furthermore,
GHG emission costs and the respective internalisation of emission tax (see Aschauer 2013, Aschauer et al. 2015) have
not been considered in this model, since LSP must not pay directly any GHG emission costs in Germany. Transport
lead time costs and truck costs per hour (see Aschauer 2013, Aschauer et al. 2015) have also been excluded, since
these cost parameters are already integrated into the variable costs.
The relationship between average shipment amount and total transport costs has not been measured for this ap-
proach. This full-cost calculation for trucks could be used for different truck types. In this study, we concentrate on
light duty vehicles (3.5 tonnes maximum permissible weight).
<Mileage
performed in line +
haul> <Road freight
Share of transport transport volume>
costs for line haul
-
<Shipment
+
amount> - Average transport
Total transport
+ costs + costs per trip
<Fixed costs>
<GHG emission <Tyre prices>
costs> <Average operating
+ <Total road mileage life of vehicles> - <Transport stock>
<Total amount of <Transport lead + performed>
time costs> + + - +
fuel consumed> + +
+ Time
Fuel costs Maintenance/repair + depreciation +
+ +
Fuel price + Variable + costs Other costs
+ costs <Purchase price +
Lubrication (catalogue price - - + Personnel
costs + deduction)> costs
<Residual +
value> Fixed costs
Toll costs + + +
+ + +
+ + + <Accumulated
<Road mileage +
Interest costs Taxes/insurances transport lead time>
performed for last
mile> Toll charge
Tyre costs Unit-of-production + +
+ depreciation +
Tyre-kilometres + Number of tyre + - <Transport stock>
sets needed -
Tyre prices + -
+ <Total road mileage
<Total road mileage Purchase price Residual performed>
performed> (catalogue price - value
deduction)
<New purchased
+
vehicles>
Fig. 8. Causal loop of freight transport cost development (Source: own diagram based on Wittenbrink 2011, Aschauer 2013, Aschauer et al. 2015,
VDI 2002)
Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094 1091
Thaller et al. / Transportation Research Procedia 00 (2017) 000–000 17
In this subsection, the traffic supply (infrastructure) and the impacts of the total transport demand on the general
traffic situation and the environment are qualitatively described. This submodule is derived from the research ap-
proaches of Lee (1997), Heimgartner (2001), Raux (2003), Raux et al. (2007), Emberger (1999), Pfaffenbichler
(2003), Bossel (2004a,b,c), Schade et al. (2000), IWW (2000), Schade (2004), Kuchenbecker (1999), Fiorello et al.
(2006), Engel (2010), Aschauer (2013), Aschauer et al. (2015), Jin et al. (2009), Wang et al. (2008), Hong et al. (2011),
Zhan et al. (2012), Ruutu (2013) as well as Han and Hayashi (2008).
The traffic supply is determined by the infrastructure quality – defined as gross fixed assets (fixed capital for
infrastructure). The infrastructure quality is influenced by the total transport volume negative inhibitory. Due to de-
clining infrastructure, the higher the total transport volume, the lower the infrastructure quality will be. A low infra-
structure quality enhances the investments in road infrastructure by public authorities, in this case municipalities.
These investments include investments in new infrastructure, in road way expansions and maintenance costs. In this
case, the level of traffic congestion effects positive reinforcing on the investments in road infrastructure of the public
authorities. Therefore, the investments result in an improvement of the infrastructure quality. The higher the infra-
structure quality and the investments in infrastructure, the higher the infrastructure capacity will be, which leads to a
lower infrastructure utilisation. Municipalities mostly have spatial and fiscal bottlenecks to expand their infrastructure
towards improved handling of transport flows. Because of these issues and the lack of public acceptance for such
projects, the expansion, new construction and maintenance of the traffic supply is not in the focus of interest of this
research approach. Furthermore, the pure effects of these projects on transport are illustrated by transport demand
models, which are more suitable for this purpose. The total transport volume (in this case: traffic intensity, transport
demand, number of trips/vehicles on the infrastructure or number of vehicles per time unit) consists of the number of
car trips (transport demand of MIT), the share of public transport and the road freight transport volume. In this study,
we do not consider the public transport. Therefore, the total transport volume positively reinforces the infrastructure
utilisation, resulting in a higher level of traffic congestion, which effects negative inhibitory on average speed on road.
Infrastructure capacity on road influences negative inhibitory the infrastructure utilisation and positive reinforces the
average speed on road. Furthermore, average speed on road affects negative inhibitory average travel time. In this
case, the average travel time is calculated by the average speed on road and the average travel distance (for freight
transport and passenger transport).
Within the quantified SD approach, the average travel distance and the average transport performance are not
considered, since the GHG emissions of MIT and of road freight transport are separately calculated (see chapters 4.1
and 4.3.5). In addition, the average travel time affects negative inhibitory the transport volume. This means that the
higher the travel time, the lower the transport volume will be. To calculate the total transport performance, the
transport volume and the average travel distance are used. Both parameters influence the total transport performance
in a positive reinforcing manner. The external effects or costs due to the traffic impact of the transport demand are
illustrated by noise and GHG emissions, land usage and macroeconomic costs by delays. The transport volume in
total and the level of traffic congestion affect positive reinforced the noise emission level. The noise impact is only
locally determined. The GHG emissions can be derived from the total transport performance and the level of traffic
congestion. Land usage results from land use for infrastructural projects, for commercial premises, or logistics facili-
ties settlements. Macroeconomic costs develop due to delays of transport users. Consequently, the macro economy
suffers a financial loss. The higher the noise emissions, GHG emissions, land usage and macroeconomic costs by
delays, the more the total external effects or costs will be. These costs do not have to be paid directly by the single
transport user, but are split between the total society by the nation state (in this case the Federal Republic of Germany)
and are covered by tax revenues. Therefore, the taxes can be increased by higher external costs. As regards the quan-
tification of the SD approach, the external costs are not modeled, since their internalisation and the temporal and
spatial assignment of these costs are very difficult. Moreover, the municipalities do not have to directly bear the
external costs themselves, which are generated in their territories.
To summarise, the quantified SD model focuses only on transport related parameters and its effects on the envi-
ronmental quality (e.g. GHG emissions). For this reason, the parameters noise emissions, land usage and macroeco-
nomic costs by delays are excluded for the mathematical formalisation. The average travel time is also not necessary
for the further modelling procedure, since the temporal characteristics of tours or mileage performed are considered
1092 Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094
18 Thaller et al. / Transportation Research Procedia 00 (2017) 000–000
in other submodules. In Fig. 9 the causal loop, which illustrates the traffic supply (infrastructure) development and
the impacts of the total transport demand on the general traffic situation and the environment, is presented.
In conclusion, in this contribution we discussed the research problem and derived concrete research objectives of
the overarching approach. Following this, the state of the art in SD modelling was briefly shown. The methodological
approach and the SD modelling procedure were then presented to develop our own SD approach. The core of this
contribution was the qualitative description of the urban freight landscape with the relevant stakeholders (e.g. private
households, retailer, shipper, logistics service provider) and their interdependencies by the qualitative SD approach,
causal loop diagramming. The qualitative SD model developed was explained and transparently illustrated by causal
loop diagrams which showed the internal structures of the submodules as well as the feedbacks between them.
The main focus was on decision processes and behaviour of the freight demand (e.g. private households, retailers,
shippers) and the resulting freight transport demand of the logistics service provider (LSP; in this case CEP service
provider), which not only affects freight transport volume and development but also the overarching traffic situation
and the environment at an urban level.
In further SD modelling steps, the mathematically formalised SD model will be parametrised with historical data
and tested in regard to its behaviour und structure sensitivity. In the quantified SD approach, the population will be
differentiated in three household groups by size with their respective consumption behaviour. The LSP fleet (CEP
service provider) only consists of light duty vehicles (3.5 tonnes maximum permissible weight). Therefore, a differ-
entiation in different vehicle types will not be necessary in this case. GHG emissions generated by the transport users
will have to be segmented in the quantified SD approach. Furthermore, future scenarios considering alternative poli-
cies and structures (developing drivers) will be implemented and simulated to assess them in comparison to the refer-
ence scenario (Business-As-Usual). The final SD approach and the future scenarios with their resulting outcomes will
be discussed and validated within expert interviews or workshops.
In this instance, the objective is that the final SD model could give qualitatively valid statements about developing
tendencies and dynamics as the effect of freight transport, economic activities and population development - as well
as environmental effects at an urban level. Through this SD approach, these developing tendencies can be illustrated,
assessed and simulated by medium- and long-term forecasts.
The overarching approach attempts to link the SD to a microscopic freight transport simulation (MFTS). MFTS is
used to generate agents in a realistic urban environment based on the aggregated results of SD. These agents carry out
their daily tour plans (LSP vehicles) and daily travels (private households, retailers). Private households and retailers
also order deliveries from the shipper. At that point, the tour plans are executed and assessed by an iterative process.
The outcome of MFTS is transferred to SD, again to mainly adjust the spatially-dependent parameters in SD, which
could not be measured in a spatial context in SD. SD is therefore a suitable methodology to map, assess and simulate
urban freight transport within medium- and long-term forecasts.
This approach serves as a moderation instrument for political and logistics decision making processes. The model-
ling framework could also be adapted to other markets or spatial levels. The functionality and structure of SD could
be the same, yet other stakeholders have to be modelled and therefore other data has to be collected.
Acknowledgements
The presented compilation and analysis were integrated in the project ´Sustainable freight transport in urban areas
– An impact assessment by System Dynamics based freight transport models´. We gratefully acknowledge the funding
of this work enabled by the German Research Foundation (DFG) under contract CL 318/15-1 and LE 2881/1-1.
References
Aschauer, G.J., 2013. The development of sustainable transport through the right logistics strategy – a system dynamics approach, 31st
International Conference of the System Dynamics Society. Cambridge, Massachusetts, USA, 21–25 July 2013.
Aschauer, G.J., Gronalt, M., Mandl, C., 2015. Modelling interrelationships between logistics and transportation operations – a system dynamics
approach. Management Research Review 38 (5), 505–539.
Balmer, M., Rieser, M., Meister, K., Charypar, D., Lefebvre, N., Nagel, K., Axhausen, K.W. (2009). MATSim-T: Architecture and Simulation
Times. In: Bazzan, A.L.C., Klügl, F., (eds.). Multi-Agent Systems for Traffic and Transportation Engineering. Information Science
Reference, Hershey, 57-78.
Bossel, H., 2004a. Systeme, Dynamik, Simulation – Modellbildung, Analyse und Simulation komplexer Systeme. Books on Demand GmbH,
Norderstedt.
Bossel, H., 2004b. Systemzoo 2, Klima, Ökosysteme und Ressourcen. Books on Demand GmbH, Norderstedt.
Bossel, H., 2004c. Systemzoo 3, Wirtschaft, Gesellschaft und Entwicklung. Books on Demand GmbH, Norderstedt.
De Jong, G., 2004. National and International Freight Transport Models. An Overview and Ideas for Future Development. Transport Reviews 24,
103–124.
Emberger, G., 1999. Interdisziplinäre Betrachtung der Auswirkungen verkehrlicher Maßnahmen auf sozioökonomische Systeme. Doctoral Thesis
1999, Institut für Verkehrsplanung und Verkehrstechnik. Vienna, Austria.
Engel, K., 2010. Demografische Determinanten der Energienachfrage der Haushalte in Deutschland. Doctoral Thesis 2010, Universität Erlangen-
Nuremberg. Erlangen, Nuremberg, Germany.
Fiorello, D., et al., 2006. Scenarios for theTransport system and Energy supply and their Potential effects STEPs – Transport strategies under the
scarcity of energy supply. Buck Consultants International, Hague.
Forrester, J.W., 1961. Industrial dynamics. Pegasus Communications, Waltham.
Gacogne, V., 2003. Impact des couts de transport sur les systèmes logistiques par une modèlisation en dynamique des systèmes – Le modèle
SANDOMA. Champs-sur-Marne, France.
Han, J., Hayashi, Y., 2008. A system dynamics model of CO2 mitigation in China’s inter-city passenger transport. Transportation Research Part
D 13, 298–305.
Heimgartner, C., 2001. Systemdynamische Simulation von Verkehr und Flächennutzungen – Evaluation nachhaltigkeitsfördernder Maßnahmen.
Institut für Verkehrsplanung, Transporttechnik, Strassen-und Eisenbahnbau. Zurich, Switzerland.
Hong, Y., Liyin, S., Yongtao, T., Jianli, H., 2011. Simulating the impacts of policy scenarios on the sustainability performance of infrastructure
projects. Automation in Construction 20, 1060–1069.
Institut für Wirtschaftspolitik und Wirtschaftsforschung (IWW), 2000. ASTRA – Assessment of Transport Strategies. Final Report 2000,
London, Great Britain.
Jin, W., Xu, L., Yang, Z., 2009. Modelling a policy making framework for urban sustainability: incorporating system dynamics into the
ecological footprint. Journal of Ecological Economics 68(12), 2938–2949.
1094 Carina Thaller et al. / Transportation Research Procedia 25 (2017) 1075–1094
20 Thaller et al. / Transportation Research Procedia 00 (2017) 000–000
Kaczmarek, M., Völker, M., 2004. Entwicklung von Simulationsmodellen für die Analyse von Supply Chain-Strategien und -Strukturen im
ProC/B-Paradigma. Technical Report 03020. Dortmund, Germany.
Kapmeier, F., 1999. Vom systemischen Denken zur Methode System Dynamics. Diploma Thesis 1999, Betriebswirtschaftliches Institut der
Universität Stuttgart, Abteilung IV, Lehrstuhl für Allgemeine Betriebswirtschaftslehre und Betriebswirtschaftliche Planung. Stuttgart,
Germany.
Kuchenbecker, K., 1999. Strategische Prognose und Bewertung von Verkehrsentwicklungen mit System Dynamics. Karlsruher Beiträge zur
wirtschaftlichen Forschung 8.
Kühn, A.; Krail, M., 2013. Dynamic demand modeling of freight fleets, 31st International Conference of the System Dynamics Society.
Cambridge, Massachusetts, USA, 21–25 July 2013.
Lee, S.Y., 1997. An integrated model of land use/transportation system performance: System dynamics modeling approach. Transportation
Research Part A 1 (31), 79.
Meyer Sanches, L., Pinto, J.A., Fontes, L.J., 2013. Freight Vehicle Circulation Restriction Policy in an Emerging Country Metropolitan Area:
undesired impacts, 31st International Conference of the System Dynamics Society. Cambridge, Massachusetts, USA, 21–25 July 2013.
Moder, N., 2010. Standard-Vorgehensweise zur Analyse und Optimierung der Distributionslogistik im Bereich Business to Consumer. Doctoral
Thesis 2010, Fakultät Maschinenbau, Technische Universität Ilmenau. Ilmenau, Germany.
Mousseau, B., Fulda, A.S., Emberger, G., 2012. Modelling impacts of local policies to optimise the development of electric vehicles, mobil.TUM
2012 – International Scientific Conference on Mobility and Transport. Munich, Germany.
OECD, 2012. OECD-Umweltausblick bis 2050. Die Konsequenzen des Nichthandelns. Zusammenfassung.
Pfaffenbichler, P., 2003. The strategic, dynamic and integrated urban land use and transport model MARS (Metropolitan Activity Relocation
Simulator) – Development, testing and application. Doctoral Thesis 2003, Institut für Verkehrsplanung und Verkehrstechnik. Vienna,
Austria.
Raux, C., 2003. A system dynamics model for the urban travel system, European Transport Conference. Strasbourg, France, 8–10 october 2003.
Raux, C., Sdika, M., Hermenier, V., 2007. Simulation de la dynamique du système des déplacements urbains: une plate-forme de modélisation.
Rapport final.
Rosenberg, R., Margolis, D., Karnopp, D., 2012. System dynamics: modelling. Simulation and control of mechatronic systems. Wiley, London.
Ruutu, S., Auvinen, H., Tuominen, A., Ahlqvist, T., Oksanen, J., 2013. Simulation of Transitions Towards Emission-Free Urban Transport, 31st
International Conference of the System Dynamics Society. Cambridge, Massachusetts, USA.
Salini, P., Karsky, M., 2002. SimTrans – Freight Transportation Simulation Model, 20th System Dynamics Conference. Palermo, Italy.
Schade, B., Rothengatter, W., Schade, W., 2000. Ecological and Economic Modelling with a Model for Economic assessment of Sustainability
policies Of Transport (ESCOT). 3rd Biennial Conference of the European Society of Ecological Economics. Vienna, Austria.
Schade, W., 2004. Assessing direct or indirect benefits of transport? Comparing benefits of transport policies within the transport market versus
within other markets with the ASTRA model, 10th World Conference on Transport Research. Instanbul, Turkey, 4–8 July 2004.
Schröder, S., Zilske, M, Liedtke, G., Nagel, K., (2012). Towards a multi-agent logistics and commercial transport model: The transport service
provider’s view. Procedia - Social and Behavioral Sciences 39, 649-663.
Statistisches Bundesamt, 2008. Verkehr. NST-2007. Einheitliches Güterverzeichnis für die Verkehrsstatistik – 2007. Wiesbaden.
Sterman, J., 2001. System Dynamics Modeling: Tools for Learning in a Complex World. California Management Review 43 (4).
Tavasszy, L.A., Halim, R.A., Seck, M.D., 2012. Modeling the Global Freight Transportation System: A Multi-Level Modeling Perspective,
Winter Simulation Conference. Berlin, Germany.
Thaller, C., Clausen, U., Kampmann, R., 2016a. System Dynamics Based, Microscopic Freight Transport Simulation for Urban Areas, in
„Commercial Transport“. In: Clausen, U., Friedrich, H., Thaller, C., Geiger, C. (Eds.). Springer International Publishing, New York, pp.
55–72.
Thaller, C., Dahmen, B., Liedtke, G., Friedrich, H., 2016b. Freight Transport Demand Modelling, in „Commercial Transport“. In: Clausen, U.,
Friedrich, H., Thaller, C., Geiger, C. (Eds.). Springer International Publishing, New York, pp. 39–54.
Verein Deutscher Ingenieure (VDI), 2002. VDI-Richtlinie 4400, Blatt 3: Logistikkennzahlen für die Distribution. Beuth Verlag GmbH, Berlin.
Villa, S., Goncalves, P., Arango, S., 2013. Exploring Retailer’s Ordering Decisions under Delays, 31st International Conference of the System
Dynamics Society; Cambridge, Massachusetts, USA, 21–25 July 2013.
Wang, J., Lu, H., Peng, H., 2008. System Dynamics Model of Urban Transportation System and Its Application. Journal of Transportation
Systems Engineering and Information Technology 8 (3), 83–89.
Weidmann, U., Stölzle, W., Bopp, B., Hofmann, E., 2012. Nachhaltige Güterfeinverteilung – Ein systematischer Ansatz. Internationales
Verkehrswesen 64 (2), 28–33.
Wittenbrink, P., 2011. Transportkostenmanagement im Straßengüterverkehr – Grundlagen, Optimierungspotenzial, Green Logistics. Gabler,
Wiesbaden.
Zhan, S.F., Zhang, X.C., Ma, C., Chen, W.P., 2012. Dynamic modelling for ecological and economic sustainability in a rapid urbanizing region.
Procedia Environment Sciences 13, 242–251.