How do train-cyclists navigate? : exploring bike-train route choice behavior in the Amsterdam Metropolitan Area

(1)

= ?

Author:

Pedro Nieves

Student number:

11659912

Email:

pgna.pr@gmail.com

Submission Date:

11 June 2018

Supervisor:

Dr. Marco te Brömmelstroet

Second reader:

Dr. Pieter Tordoir

Master’s program:

Urban & Regional Planning

How do train-cyclists

navigate

_?

Exploring bike-train route choice behavior in the

Amsterdam Metropolitan Area

(2)

2

University of Amsterdam

Faculty of Social and Behavioral Sciences

Graduate School of Social Sciences

How do train-cyclists navigate?

Exploring bike-train route choices in the

Amsterdam Metropolitan Area

Supervised Research Project

Written by Pedro Nieves

In partial fulfillment of the graduation requirements for the

Master of Science in Urban and Regional Planning

(3)

3

Acknowledgments

First and foremost, I would like to thank my supervisor, Marco te Brömmelstroet (a.k.a. the

fietsprofessor), for his guidance, inspiration, and constant feedback throughout my research. I

am extremely grateful for his availability, attentiveness, and for his constructive criticism which

challenged me to think outside of the box, encouraging me to write a better thesis within the

timeframe.

I would also like to thank my colleagues from the University of Amsterdam with which I had the

chance to discuss my study. I really appreciated their personal support and remarks throughout

these past six months. Their friendship, cheerfulness, and dedication have inspired me and made

this an unforgettable experience.

The past months living in Amsterdam have been very challenging yet rewarding – I came out of

this program as a different person, feeling inspired and ready to contribute towards a better

future in my home(is)land, Puerto Rico.

Finally, I want to express my deepest gratitude to my friends and family back home for their

unconditional love, support, and comfort during my studies. Special thanks to my sister,

Alexandra, and to my parents who have always been there for me throughout my academic

career. I share this achievement with you. Gracias por tanto.

Sincerely,

_______________________

Pedro Nieves

(4)

4

Abstract

Route choice studies have given attention to the factors that influence cyclists and transit users’ route decision making while disregarding the actual process of selecting a particular route from a different set of routes. This thesis has further explored this process by understanding the bicycle-train system as a distinct mode of transportation because it left open the question of how ‘train-cyclists’ make route choices, particularly from the complexity of train stations’ overlapping catchment areas. The study area which was chosen to delve into this aspect was the Amsterdam Metropolitan Area in the Netherlands because of its highly developed and integrated bicycle and railway infrastructure. Results indicate that bike-train route options increase exponentially when traveling to or from overlapping areas. All-in-all, the main finding suggest that train-cyclists are willing to have longer cycling journeys in order to shorten their train journeys.

Keywords: mobility, integrated transport, bike-train route choice, urban cycling, rail transport

Resumen

Los estudios de elección de ruta han prestado atención a los factores que influyen la selección ruta tanto de ciclistas y como de los usuarios de tránsito sin profundizar el proceso que involucra seleccionar una ruta particular sobre un conjunto de diferentes rutas. Esta tesis ha explorado este proceso conceptualizando el sistema bici-tren como un modo distinto de transporte debido a que dejó abierta la pregunta de cómo los/as ‘ciclistas-de-tren’ toman sus decisiones de ruta, particularmente desde la complejidad de las áreas de cobertura superpuestas de las estaciones de tren. El área de estudio que se eligió para explorar este aspecto fue el Área Metropolitana de Ámsterdam en los Países Bajos debido a que su infraestructura de ciclismo y ferroviaria está altamente desarrollada e integrada. Los resultados indican que las opciones de ruta aumentan exponencialmente cuando se viaja hacia o desde áreas superpuestas. En general, el principal hallazgo sugiere que los/as ciclistas-de-tren están dispuestos a tener viajes más largos en bicicleta para acortar el tiempo de viaje en tren.

(5)

5

1. Introduction

Route choices are an outcome of our everyday mobility needs. Either going to work or meeting with a friend, every decision we make to engage in society entails a route choice. In most industrialized countries, people are becoming more and more dependent on mobility in order to participate in social and economic life (Bertolini, 2012). Unfortunately, the current mobility system is dominated by motorized means of transport which contributes to a large amount of CO₂ emissions (Faiz et al. 1996) and thus compromises the public health of future generations. Cities have therefore begun to take measures for the sake of transition towards a more sustainable urban mobility system.

The integration of the bicycle with transit systems is considered as one of the utmost tactics for competing with unsustainable motorized modes of transport (Rietveld, 2000; Brons et al.,2009). Luckily, there is an increasing international trend for developing public transportation in conjunction with the bicycle (Pucher & Buehler, 2012). Bike-and-ride1_{trips provide several social and environmental benefits such as reducing}

air and noise pollution as well as congestion-mitigation benefits for communities (Martens, 2004; Krizek & Stonebraker, 2010). Various scholars have focused on the added value that the bike provides as a ‘feeder’ mode to public transport (e.g. Martens, 2007; Pucher & Buehler, 2009; Cervero et al., 2013; Halldórsdóttir et al., 2017), however, none of these studies accounted for the distinct characteristics that arise by the incorporation of the bicycle with transit systems.

On the contrary, Kager et al. (2016) addressed this matter by proposing the combined use of cycling and transit from a holistic approach stating that both systems ought to be considered as a particular mode of transport instead of two independent transport modes. This mode was conceptualized as the ‘bicycle-train system’ with regards to the underlying, distinct mechanism that it provides. Consequently, in order to upscale the usage of this system, research has given attention to modeling the route choices for both cyclists and transit users. Nonetheless, there is a limited amount of studies in which route choices are modeled for bike-train journeys (e.g. Fiorenzo-Catalano et al., 2004; La Paix & Geurs, 2015).

More significantly, there is also a lack of knowledge of how and why do bike-train users2_{make route}

choices in accordance with the synergies caused by the bicycle-train combination. In a context with a highly developed bike-train infrastructure, the speed and spatial reach of the train combined with the flexibility and high accessibility of the bicycle has direct implications on train stations, not only extending their coverage (Advani & Tiwari, 2006) but also creating overlaps of the catchment areas (Kager & Harms, 2017) in which users can have more than one station to select from (Kager et al., 2016). Understanding

1_{According to Martens (2007), bike-and-ride is a form of mobility which alludes to the joint use of the bicycle with public transportation} (e.g. trains and buses).

2_{For the purposes of this thesis, bike-train users or bike-and-ride commuters (i.e. people that make a trip using the bicycle and the train} in combination) will also be regarded as train-cyclists.

(8)

8

this from the perspective of route choice is crucial because it can potentially reveal unknown patterns of travel behavior which can inform the transport planning practice. From a policy perspective, this research could contribute to the fields of integrated and sustainable transportation by encouraging an integration of urban planning and operation of both the bicycle and the train system with the development of effective bike-train policy initiatives. Lastly, the research findings could potentially encourage the idea of creating a route choice model of the bike-train system which is currently nonexistent.

The ultimate goal of this study is to inform the big theoretical debate on how do people make route choices by zooming in into the bike-train system. Such a system can help to amplify the different conceptions related to how people move within cities and around city-regions. In addition, the main objective is to enhance the understanding of factors that influence train-cyclists route choice behavior in order to better cater for their (travel) needs. Finally, it is also important to highlight that route choice has not been studied in a context where many available options are as the bike-train mode allows. Thus, the following thesis will further explore this by taking the Amsterdam Metropolitan Area in The Netherlands (NL) as a study area because of its relative maturity in bicycle-train integration. This will be done by applying a mixed methods research design addressing the following research question: How do

(9)

9

2. Theoretical Framework

This chapter outlines the theories and concepts which underlie the main research question of the thesis. Section 2.1 provides an overview of the bike-train system and its distinct characteristics. Section 2.2 discusses route choice behavior theory followed by an extensive literature review that was conducted in order to understand the state of knowledge concerning both cycling and transit route choice studies. Lastly, section 2.3 exhibits a preliminary conceptual scheme which was created based on the findings of the route choice studies.

2.1 Bike-train system

Transportation is a system of utter importance for improving society’s welfare and wellbeing (Bertolini, 2012). As Meyer and Miller (2013:2) note, a system can be defined as “a group of interdependent and interrelated components that form a complex and unified whole intended to serve some purpose through the performance of its interacting parts.” This overarching definition was applied to transport systems in general in order to address their different performance characteristics based on travel speed and level of accessibility. The concept of speed indicates the amount of space that can be traveled in a particular amount of time whereas accessibility refers to the capacity an individual has to reach different locations and accomplish a social and/or economic activity (Rodrigue et al., 2006; Ferreira et al., 2017). When it comes to transport planning, both concepts are considered to be the most significant performance measures in determining the competence of a system for providing opportunities for mobility (Bertolini & le Clercq, 2003; Meyer & Miller, 2013).

From this system perspective, Kager et al. (2016) coined the term ‘bicycle-train system’ to distinguish a particular mode of transport whom performance integrates both the flexibility and high accessibility of the bicycle with the high speed and spatial reach of the transit system (Figure 1). ‘Train’ in this concept refers too high capacity transit services as defined by Meyer and Miller (2013:10) which includes rapid transit services that carry a high volume of passengers and whom networks connect central cities with regional centers. Thus, the concept excludes ‘feeder’ transit services such as trams and metro which generally have a lower speed. Therefore the differentiation between train versus ‘feeder’ services is based on whether a particular transit system is capable to amplify the traits of the bicycle at the level of a trip chain (Kager et al., 2016:210).

Furthermore, a journey made by the bicycle-train mode has three trip segments: (1) the home-end trip (or access travel) between the origin and the access station, (2) the train trip (or ‘main travel’) between the access and egress station, and (3) the activity-end trip (or egress travel) between the egress station and the destination (Ibid., p. 210). According to this, a bike-train trip was conditioned to whether or not it includes a combination of one or more trips made by train as part of the main travel and to whether or

(10)

10

not the home-end and/or activity-end trip is comprised of a cycling journey (Ibid., p. 210). Overall, this modal integration creates a powerful symbiotic relationship in which the bicycle and the train benefit from each other’s supply components.

Figure 1: The bicycle-train mode based on travel speed and level of accessibility as defined

by Kager et al. (2016) and adapted from Meyer & Miller (2013).

Perhaps the greatest potential that the bicycle-train has is related to the synergetic effects that it provides to its users, a phenomenon that has been disregarded in the literature. Although Pucher and Buehler (2009:101) and Krizek & Stonebraker (2010:166) acknowledged that the integration of cycling with public transport creates synergies, they did not further elaborate on how they manifest in reality. On the contrary, Kager et al. (2016:212) outlined the following modal performance dimensions as potential bike-train synergies3_:

(1) Speed and spatial reach: The high speed and spatial reach of the train compensate for the low speed and reach of the bicycle as people tend to cycle less as travel distances increase (Broach et

al., 2012; Ton et al. 2017).

(2) Adaptability: Cycling is a highly individualized means of mobility which offers door-to-door accessibility and more flexibility in departure time, speed, and route choice whereas railway transport is spatially confined to train cars.

(3) Activity chain: Traveling by train provides the opportunity to fulfill other activities (e.g. reading, working) while cycling is confined to act of mobility. On the contrary, cycling itself is an activity that offers physical and mental health benefits (Garrard et al. 2012).

(11)

11

All these characteristics are precisely what makes the bike-train a distinct mode of transportation. In accordance with the first synergy, the speed and spatial reach of the bicycle has direct implications on train stations by extending their coverage area4_{(Kager & Harms, 2017). This in turn creates overlaps}

between the stations’ catchment area in which train-cyclists may choose between more than one station according to a variety of reasons which can be related to either the services and characteristics that a station offers (La Paix & Geurs, 2015), to individual preferences (Meyer & Miller, 2013; Kager et al., 2016) and/or as a response to unexpected mishaps (Halldórsdóttir et al., 2017) such as weather or disruptions in the system5_{. By implication, the overlap of stations leads to overlapping routes which increases the}

variety of options to pick from when traveling to or from a station (Hoogendoorn-Lanser & Bovy, 2007). Bike-train synergies can thus help to understand train-cyclists’ route choice behavior by looking deeper into the complexity of catchment areas due to the overlaps of train stations.

2.2 Route choice behavior

Understanding route choice behavior is pivotal to disclose travelers’ preferences in order to improve the predictive power of traffic forecasting models (Hood et al., 2011; Anderson et al., 2017). The process of choosing a specific route for a journey from a diverse set of routes is commonly regarded by scholars as a ‘black box’ (Figure 2), that is, as a complex travel

phenomenon whose underlying mechanisms are not readily understood (Bovy & Stern, 1990; Chang & Chen, 1995). In order to address this, Bovy and Stern (1990:30) assert that individual route choice is a rational behavior mainly related to the following statements:

• the traveler, with his/her subjective needs, experiences, preferences, perceptions, etc. • the physical environment, with its objective

opportunities and their characteristics

In the case of the former, a subjective environment is highlighted as a domain which has an influence over how people navigate. This is partly determined by travelers’ socio-demographic backgrounds such as their sex, age, and employment status (Antipova et al., 2011). Nevertheless, route selection is also understood

4_{The catchment area of train stations is generally defined based on a walking distance of approximately 0.5 to 1 km (Daniels & Mulley,} 2013). On the contrary, cycling can cover up to three times this distance (Kager & Harms, 2017) which by implication increases stations’ catchment area.

5_{The study of how disruptions in the system influence bike-train route choice is beyond the scope of this paper.}

Figure 2: A general scheme of route selection as proposed

(12)

12

as a personal matter in which individual variations in terms of travel preferences6_{will befall and thus}

cannot be strictly reduced to observable personal characteristics (Bovy & Stern, 1990). On the other hand, travelers’ prior experiences also play an important role when choosing a particular (set of) route(s). The feedback gained from the usage of previously selected routes between origins to destinations informs a traveler about the different alternatives that he/she has at his/her disposition. In turn, the traveler gains a certain route knowledge by learning about the routes’ “relevant attributes that influence [his/her] trade-off and choice” (Ibid. p. 32).

However, the traveler in question will not necessarily know the total amount of available route options he/she has whilst the ones that are known will not always be considered as genuine alternatives (Xu et

al., 2011). Here is where the physical environment comes into to the picture which includes the built-up

environs, the transport network infrastructure, and contextual occurrences (Bovy & Stern, 1990). After the traveler learns the different set of routes that he/she is able to choose, a ‘factor-importance-hierarchy’7_{is made based on their various network characteristics in order to filter}8_{them and therefore}

decide which are more suitable and convenient for his/her journey (Ibid.). In this sense, travelers’ route choice behavior with regards to their subjective needs, experiences, and preferences is derived from the physical environment and its objective opportunities.

Furthermore, route choice models9_{have been created in order to predict the flows on specific links of a}

transport network by identifying the attributes that influence observed behavior and therefore determine the relationship between route choice behavior and explanatory attributes (Ibid.). Among these models, logit and probit approaches10_{are perhaps the most commonly used to predict the conditional probability}

of selecting a route from an identified choice set of origin-destination routes. The underlying logic of most available route choice models has its grounds on random utility theory11_{which is based on the assumption}

that every person is a “rational decision-maker, maximizing utility relative to his or her choices”12

(Cascetta, 2009).

6_{Travelers’ preferences are to some degree contingent upon their purpose for doing the journey (e.g. going to work or for leisure).} 7_{According to Bovy and Stern (1990), this term refers to a route choice process in which travelers rank their route alternatives taking} into consideration their network characteristics in order to determine which route(s) he/she will select. Likewise, this hierarchy is considered to be the end state of travel decisions which comes after travelers’ destination choice and mode choice.

8_{This filtering procedure has two dimensions: first, there is a perception filter in which the traveler has some degree of awareness of the} existing route alternatives and their characteristics; and second, there is an evaluation filter in which a trade-off is made based on these perceptions which are then converted into a desirability scale (Bovy & Stern, 1990:33).

9_{Some route choice models are also labeled as ‘discrete-choice models’ (Bovy & Stern, 1990).}

10_{Logit and probit models are both statistical models which account for behavioral variations such as route choice and mode choice.} 11_{This theory goes by the hand with the popular logic between transport economists in which travel time is regularly understood as} wasted time based on financial terms.

12_{The conception that travelers generally tend to behave rationally has been confronted by a handful of transport scholars whom} implicitly argue that irrationality is also constitutive of travel behavior (e.g. Bonsall & Cho, 2000; Avineri & Prashker, 2006; Xu et al., 2011).

(13)

13

The following subsections will discuss the outcomes of several bicycle and transit route choice studies in order to identify the route-related attributes which were used for predicting route choice behavior. Appendix A and B exhibits an overview of all these studies.

2.2.1 Bicycle route choice

There is a great diversity of factors that are commonly used for modeling bicycle route choice which can be grouped into socio-demographic and network attributes. In the case of the latter, cyclists’ sensitivity towards motorized traffic volume and steep hills (slope) has been found which indicates that they will tend to favor routes that have a flat topography and with a small amount of motorized vehicles passing by (e.g. Sener et al., 2009; Menghini et al., 2010; Hood et al., 2011; Beheshtitabar et al., 2014). Additionally, the findings presented by Broach et al. (2012) suggests that routes with fewer stop signs and traffic lights were preferred by cyclists, that they were also sensitive to travel distance, and that bike boulevards have a high value to cyclists during their journey. Furthermore, other studies have found how roadways’ speed limit and the absence or presence of a bike lane are likely to have a negative impact in bicycle route decision making (Hunt & Alberta, 2001; Casello & Usyukov, 2014).

On the contrary, in a recent study conducted by Ton et al. (2017), cyclists were found to be insensitive to separated bicycle paths and to motorized traffic yet readily affected towards contextual attributes such as rain and sunset/sunrise times. In addition, their results indicate that cyclists tend to minimize travel distance and the number of intersections per kilometer. On the other hand, cyclists’ characteristics such as age, gender/sex13_{, and cycling experience were used in some of these studies although their results}

suggest that these attributes have a lower impact upon bicycle route choice behavior as opposed to network attributes. Lastly, most of these studies strongly suggested either explicitly or implicitly that travel time is the most important attribute having an influence upon bicycle route choice behavior. Table 1 exhibits an overview of all the attributes that were identified as the most influential and the ones which were also repeated in the studies. 14

Table 1: Potential determinants of cyclists’ route choice behavior, based on findings of the above-cited studies.

13_{Some studies did not differentiate between these two concepts, however, there is a distinction. When speaking of gender, it refers to} the socially-constructed subjectivities used to express ones’ identity whereas sex expresses the physiological and biological differences (internal and external) that structure human corporeality (see Butler, 1990; Wittig, 1993; Preciado, 2002).

(14)

14

2.2.2 Transit route choice

Previous research has identified a fair amount of attributes that may be influencing transit route choice behavior which can also be grouped into socio-demographic, network, and contextual attributes. Gentile

et al. (2005) used line waiting times and general headway distribution as route choice determinants for

transit users. Their findings indicate that the availability of online information determines the probability of boarding a particular train at a station. Hoogendoorn-Lanser and Bovy (2007) estimated multimodal route choices considering a wide range of attributes for the three trip segments of train travel. Their main results suggest that overlaps in train journeys (the ‘main travel’) are valued positively which thus indicates that overlapping train routes are considered to be more attractive.

In another study conducted by Eluru et al. (2012), network attributes such as the number of transfers and waiting times as well as socio-demographic attributes were used to model transit route choices. The findings suggest that train users tend to favor fewer transfers which also have lower waiting time. However, as age increases, there appears to be less sensitivity towards waiting times. Moreover, Brands

et al. (2014) study focused on network attributes taking into consideration the type of station, number of

stops, and vehicle’s departure time. The general results insinuate that travel time is the principal factor influencing transit route choice. On the other hand, La Paix and Geurs (2015) study offer some interesting insights with regards to station choice. Their findings inform that the provision of unguarded bicycle parking facilities in stations may encourage travelers to reach a specific transit station and that car availability negatively influences a bike-train combination.

Other network attributes such as in-vehicle travel time and transfer times have also been found to be important factors determining transit route choice (e.g. Brahmaiah et al., 2017; Anderson et al., 2017) Lastly, Xu et al. (2018) focused on network attributes including in-vehicle crowding as a contextual attribute. Their findings suggest that crowding is also another important factor influencing route selection because it causes train delays and thus increases either in-vehicle travel time or transfer times. All-in-all, the majority of these transit route choice studies suggest that train commuters are inclined to select route alternatives which minimize their travel time. Table 2 shows an overview of the most influencing transit route choice attributes which were considered in these studies

(15)

15

2.3 Conceptual framework (preliminary)

The scheme below (Figure 3) was created to represent the preliminary conceptual framework of this thesis. The main objective of this scheme is to provide an overview of latent route choice attributes from the perspective of the bike-train system which altogether may be influencing train-cyclists route choice behavior. Further on, all bicycle and transit attributes consider in this scheme will be filtered in order to obtain a final conceptual scheme of bike-train route choice.

Route choice scholarship suggests that diverging factors are known to be influencing bicycle and transit route choices. In case of the former, network and contextual attributes appeared to be having a greater impact upon cyclists such as traffic volume, presence or absence of a bike lane, slope, and weather conditions. On the other hand, transit users’ route choices seem to be influenced mainly by network attributes such as the number of transfers and the type of station. For the socio-demographic attributes, they were the least ones suggesting a significant influence according to the overall results. More significantly, the findings from nearly all the route choice studies that were covered indicated that travel time is the utmost factor which plays a role in route decision making.

(16)

16

3. Methodology

This chapter aims at outlining the systematic phases undertaken to conduct this research. Section 3.1 starts off by presenting the problem guiding the main objective of the thesis. Section 3.2 shows the main research question followed by four sub-questions whereas section 3.3 briefly discusses the study area selected for this study. Finally, section 3.4 exhibits the chosen research design and explains the methods chosen to assess each of the research sub-questions.

3.1 Problem Statement

Several studies have tried to explain route choices by focusing on socio-demographic, network, and contextual characteristics of hypothetical or observed routes while disregarding the actual process of selecting a particular route. This ‘black box’ requires more theoretical development relating to its inner performance (Bovy & Stern, 1990) which becomes even more complex when there are overlaps between train stations’ catchment area. The phenomenon of overlapping stations has far-reaching implications for train-cyclists’ travel behavior because it increases the availability of routes to choose from (Kager & Harms, 2017) and, by implication, increases the capacity to personalize the whole journey based on various route-related attributes. Nevertheless, none of the studies mentioned above referred to the overlap of stations’ catchment area as an indicator that has an impact on either bicycle or transit route choice. There is therefore a lack of knowledge regarding how these overlaps are actually having an influence on route choice behavior.

By the time of writing, research on bike-transit route choice was found to be very limited. The literature clearly evidenced that bicycle and transit route choices are often modeled separately. In the same manner, multi-modal route choices have also been modeled but not exclusively for the bicycle-train system which by definition considers a journey made by bike and train as a single mode of transport. Even though Brands et al. (2014) examined multiple routing considering the bicycle as a feeder to transit systems, their route choice model in the end disregarded the bicycle because they focused only on interregional journeys. On the other hand, La Paix and Geurs (2015) model acknowledge bike-train share but only to estimate mode choice rather than route choice15_.

Furthermore, in current route choice scholarship, travel time is widely considered as one, if not the most, significant factor influencing route choice behavior (Sun & Zu, 2012), especially during peak hours (Raveau

et al., 2014; Ton et al., 2017). The theoretical insights suggested by the findings of the above-outlined

route choice studies highlighted that it is usually a matter of the fastest route. However, when considering the complexity of catchment areas due to station overlaps, travel time may not be the greatest determinant for train-cyclists because of the variety of bike-train routes and the diversity of their

15_{Although their study focused on mode choice, the route-related attributes that were used and their derived results exhibit crucial} information that could inform potential bike-train route choices.

(17)

17

attributes. Therefore, understanding route choices within the complexity of the bike-train system seems to be both an interesting and urgent next line of research.

3.2 Research Question(s)

In order to address the above problem statement, it is necessary to explore travelers’ process of making bike-train route choices with regards to the immense diversity of route-related attributes. Thus, the research question of this thesis is:

How do train-cyclists make route choices while using the bike-train system?

To answer this main question, the following sub-questions were drafted: (1) How does the current bike-train system perform?

(2) What are the (in)direct attributes which influence potential bike-train route choices?

(3) To what extent are train-cyclists willing to trade-off travel time with other bike-train route choice attributes?

(4) How are bike-train route-related attributes playing a role in train-cyclists’ route choice behavior? These research questions are outlined in such a way to render a fruitful understanding of the behavioral complexity concerning train-cyclists’ route choices based on the theoretical underpinnings of the bike-train system. Likewise, having four sub-questions was chosen in order to explore the full range of the main research question considering how there is a lack of knowledge concerning bike-train route choices which are made in overlapping catchment areas.

3.3 Study Area

The Amsterdam Metropolitan Region (AMA; Figure 4) is a city-region16_{surrounding the city of Amsterdam,}

the capital of the NL. It is comprised of 33 municipalities and extends over two Dutch provinces composed of North Holland and Flevoland. Approximately 2.2 million people reside within this area which accounts for over 11 percent of the total Dutch population. The AMA is also the NL’s most robust economic region playing a central role in the international market. Furthermore, there are a total of 52 Nederlandse

Spoorwegensystem (NS) railway stations within the AMA’s administrative boundaries of which inhabitants

can use to easily move between municipalities. This amount accounts for over 13 percent of the total amount of train stations in the NL. Additionally, only 20 municipalities from the AMA currently have stations within their territories.

16 _{This concept is a political and analytical transition towards a new regionalism, coupled to encourage a re-escalation of State} intervention. A city-region is generally defined based on adjacent built-up areas with high shares of economic and traffic flows (see Davoudi, 2009).

(18)

18

Figure 4: Study area of the thesis. Source: Author’s illustration.

Since the beginning of the 1990s, the Dutch national government has dedicated plenty of its investments towards the integration of the bicycle with the NS railway system (Martens, 2007). Of all the 17 million inhabitants of the NL, it is estimated that around 1.2 million people are daily train commuters of which 47 percent use the bicycle to access stations and around 12 percent cycle from the egress stations (Kager et

al., 2016:208). By the time of writing, there are a total of 398 railway stations17_{in the NL and it is estimated}

that 69 percent of the Dutch population resides within a 5 km radius of the stations in cycling distance. Also within this reach, there are usually two or more stations of which to choose from which are interconnected with the bicycle network (Ibid., p. 213-214). Overall, this data strongly suggests why the Dutch have a high bike-train ridership. Given these points, the AMA was selected as a study area because of its relative maturity in bicycle-train integration.

(19)

19

3.4 Research Design

The empirical strategy chosen for this thesis was a mixed methods research approach. As Bryman (2012:628) argues, this research strategy is useful because it yields the combination of both quantitative and qualitative methods in such a way that “would seem to allow the various strengths to be capitalized upon and the weaknesses offset somewhat.” The strategy was selected because it allowed triangulating the data by, on the one hand, acquiring quantitative data by conducting web-based surveys but also, on the other hand, by having qualitative insights from interviews as a data source. The following sub-sections were divided per research sub-question explaining their methods of both data collection and analysis. Table 3 shows the summary of the research design selected for this study.

Table 3: Research design for this thesis.

3.4.1 How does the current bike-train system perform?

In order to have a sound knowledge of the bike-train mode, it is necessary to illustrate its current system performance and the route opportunities it offers to its users. Therefore, the units of analysis for this research question were the NS railway stations and potential bike-train route choices.

To answer this sub-question, two phases were undergone. Firstly, Esri’s cartographic software known as Geographic Information Systems (GIS) was used as a research tool18_{. According to Lejano (2008), some of}

the traits of using GIS are its potential to uncover problems that arise in the planning profession and its capacity to visualize reality by mapping a large number of datasets. This software was thus chosen because it enabled to easily illustrate the system performance of the bike-train mode within the AMA. The data of the NS railway system required to answer this question was collected from a GIS open-data platform of a Dutch organization called Imergis19_{which provides a wide variety of data sets of the Netherlands. The}

data was then analyzed by conducting a density, an overlay, and a proximity analysis using various

18_{https://www.esri.com/en-us/what-is-gis/overview} 19_{http://www.imergis.nl/asp/47.asp}

(20)

20

geoprocessing tools20_{which results where then cartographically represented. The resulting maps were}

then used for the empirical strategy of the last research sub-question.

Thereafter, during the second phase of this question, three hypothetical travel situations were determined for train-cyclists that live in different cities of the AMA. This was done by estimating how much travel time did it take train-cyclists to get from their origins to their destinations. Data of both home-end and activity-home-end trips (cycling routes) was appraised using Google Maps™’ route planning service21_as

a tool and its embedded algorithm which represents routes graphically and provides an estimation of the travel time per routes available that each traveler has at his or her disposal. On the other hand, for the main travel (train routes), the NS’s journey planner service22_{was used to determine the fastest train trips.}

These two tools were chosen because they provided real-time travel data associated with rail tracks, roads, and traffic restrictions, therefore reassuring the validity of the travel scenarios (Santos et al., 2011; van Veenendaal, 1993). Afterward, both ArcGIS for Desktop and ArcGIS Online were used to map all the routes that each train-cyclist had available per trip segment.

The travel times for each trip segment were expressed in minutes and were then aggregated to display the totality of bike-train route options. In sum, the general aim of this question was to set the scene for the bike-train mode by exhibiting its system performance and by showing the number of similar route choices a train-cyclist has in terms of the travel time that he/she incurs when traveling from point A to point B.

3.4.2 What are the (in)direct attributes which influence potential bike-train route choices?

Travelers route decision making is contingent upon a diverse set of route-related attributes which are (in)directly considered based on their preferences and needs (Bovy & Stern, 1990). Consequently, the units of analysis for this sub-question were the bicycle and transit route choice attributes outlined in the preliminary conceptual scheme (Figure 3). The main objective was to filter these attributes by conducting a web-based expert survey (see Appendix C) as a secondary data

gathering method using the free online platform called Google Forms23_{. This method was chosen because it proved to be a fast}

and convenient strategy to acquire data from experts in a short amount of time considering their tight schedules. Furthermore, the experts invited to participate were those well-acquainted with bike-train research and developments in the NL. By using a Likert scale24_{(Table 4), these experts were asked to rank each of}

20_{http://desktop.arcgis.com/en/arcmap/10.3/main/analyze/geoprocessing-tools.htm} 21_{https://www.google.com/maps}

22_{https://www.ns.nl/en/journeyplanner}

23_{https://www.google.com/forms/about/}

24_{The Likert scale used in this study had seven values based on levels of importance as suggested by Vagias (2006). A scale of 7 was}

selected because it allows for better data distribution in the case of small samples (Finstad, 2010).

(21)

21

the route choice attributes in terms of importance based on their influence upon train-cyclists’ route choice behavior. A Likert scale was chosen because it is useful for revealing different degrees of opinion (Joshi et al., 2015) with regard to a specific topic, in this case, for route choice. Table 5 shows the list of the experts and their respective email addresses.

Luca Bertolini is a professor of urban and regional planning and director of the Centre for Urban Studies at the University of Amsterdam. Both his research and teaching are mainly focused on the integration of transport and land-use planning and on different ways for improving the interaction between theory and practice. Wouter de Koning, current head of Mobility Services for the NS, is responsible for the operation and development for all the NS stations in the NL. He oversees different tasks such as the management of bike parking at stations as well as the management of the ‘OV-fiets’25_{. Roland Kager is a data analyst of}

mobility, land-use, cycling, and transit at a Dutch company called Studio Bereikbaar26_{. His work mainly}

focuses on tracking and monitoring travel behavior from the combined use of bike and train. Likewise, he and Luca Bertolini are co-authors of a paper27_{in which they coined the concept of ‘bicycle-train system’.}

Lucas Harms is a senior researcher at the Kennisinstituut voor Mobiliteitsbeleid (KiM; Netherlands Institute for Transport Policy Analysis) working in the field of transport and mobility. He is also co-writer of a paper28

with Roland Kager in which they provided an overview of the synergies that arise from the integration of cycling with transit. Niels van Oort works as an assistant professor at Delft University of Technology and he is also a public transport consultant at a Dutch consultancy company called Goudappel Coffeng. His main area of expertise revolves around public transport planning. Finally, Olaf Jonkeren is also a researcher at the KiM working with urban transportation in fields related to freight, maritime, and inland waterway transport. He was also the leader of a project that studied bike-train users in the NL.

Table 5: List of experts with their respective emails.

After collecting all the data from the web survey, it was then analyzed using descriptive statistics with the help of Microsoft Excel in order to filter the route choice attributes based on their rankings. This method of analysis was selected because it allowed determining which attributes were considered most important according to their high percentage of responses and to their average values. Afterward, Tableau Software

25_{The OV-fiets is a Dutch bicycle rental program available in most of the NS train stations.}

26_{https://www.studiobereikbaar.nl/}

27_{Kager et al. (2016), Characterisation of and reflections on the synergy of bicycles and public transport, Transportation Research Part}

A, Volume 85, pp. 208-219.

28_{Kager, R. and Harms, L. (2017) Synergies from improved bicycle-transit integration: Towards an integrated urban mobility system.}

(22)

22

was used to create graphs representing the results of the analyses. Lastly, at the end of the survey, experts were asked to mention the existence of other underlying attributes that were not included in the preliminary conceptual scheme. This new set of bike-train route choice attributes was added to the list. As a result, a final conceptual scheme was created based on all these attributes which were then used in the empirical strategy of the remaining sub-questions.

3.4.3 To what extent are train-cyclists willing to trade-off travel time with other bike-train route

choice attributes?

Route choice scholarship suggests that travel time has the utmost influence on travelers as opposed to other route-related attributes. Therefore, in order to understand how these other bike-train route attributes may be exchanged with travel time, the units of analyses chosen to assess this sub-question were train-cyclists. These users were those who live, work, and/or study in the AMA which travel using the bicycle to reach an access station and/or to travel from an egress station, and whom main travel is composed of a train journey using the NS railway system.

To collect the quantitative data required to answer this question, Google Forms was also used to create another web-based survey (see Appendix D). This method of data collection was selected because one of the advantages of web surveys are their ability to filter questions in order to better target respondents (Bryman, 2012). In turn, this proved to be useful by assuring that respondents complied with the above-stated conditions and therefore excluding those who did not bear a likeness to the type of traveler using the bike-train mode. Likewise, respondents were recruited using a virtual snowball sampling technique (Baltar & Brunet, 2012) by posting the link of the survey on Facebook and Twitter and by distributing small papers with the link of the survey to cyclists who accessed or egressed from Amsterdam Lelylaan station and Amsterdam Centraal station29_{. The survey was open for responses for a period of one month, from}

the 11th April 2018 until the 11th May 2018. To incentivize responses, participants were offered the opportunity to win one of three €10 e-gift cards for any purchase on Amazon.com.

Furthermore, the web survey was structured in two sections. The first part had different questions which were drafted according to the socio-demographic route-related attributes whereas the second part had 14 statements with a ‘fill-in the blanks’ format in which respondents had to choose how much travel time between 0 to 30 minutes were they willing to cycle in order to compensate other bike-train route choice attributes. Additionally, the survey filtered responses according to the type of bicycle30_{that participants}

frequently used. This was done as a strategy to differentiate the trade-offs between bike-train users.

29_{These stations were chosen because of their distinct characteristics. Amsterdam Lelylaan is a local station which only offers sprinter} services whereas Amsterdam Centraal is an intercity station which also has different facilities. Also, these stations were selected because they were the ones that I frequently visited while conducting this study.

30_{The following types of bicycle were considered: ordinary bike, e-bike, OV bike (OV-fiets), folding bike (vouwfiets), cargo bike (bakfiets),} or a swap bike (swapfiets).

(23)

23

After collecting all the secondary data, it was analyzed using descriptive statistics by showing the percentage of responses for each of the socio-demographic questions and by exhibiting the average values of the travel time that train-cyclists were willing to trade-off when making their route choices, whilst also showing the percentage of responses per each attribute. A trade-off analysis was chosen because it allowed to recognize which attributes were most valued by respondents by using a discrete choice approach (McCullough, 1998). In sum, the main purpose of this sub-question was to show the hierarchy between the attributes, that is, to see how train-cyclists rank different bike-train route attributes based on their willingness to cycle an extra amount of time. The assumptions derived from these results were triangulated with the insights provided from interviews of the following research sub-question.

3.4.4 How are bike-train route-related attributes playing a role in train-cyclists’ route choice

behavior?

The units of analysis chosen to assess this question were also bike-train users. In order to gather the qualitative data required to generate accounts of route choice, a total of five semi-structure key-informant interviews (see Appendix E) were conducted for people that lived, worked, and/or studied at the AMA. Participants were recruited using a snowball sampling technique by contacting respondents from the trade-off survey who provided their email addresses because of their interest to participate in a follow-up interview. The selection of particular interviewees was based on maximizing the diversity of train-cyclists’ socio-demographic backgrounds as a means to understand different route decision scenarios (Bryman, 2012). They were also selected because all of them frequently traveled to or from overlapping catchment areas. Likewise, pseudonyms31_{were adopted to preserve the anonymity of all participants and}

to follow the line of reasoning used to explain their travel experiences.

Furthermore, these in-depth interviews covered a variety of open-ended questions which were structured on the basis of the findings derived from the previous research sub-questions to tighten their link. During the first phase, participants were asked about the answers they provided for the questions related to their socio-demographic attributes. For the second phase, participants were shown the different maps created to illustrate the bike-train’s system performance and had to reflect upon them based on their previous travel experiences. The third and final phase contained several questions in which interviewees had to reminisce their route choices according to each bike-train route-related attribute. In general, the decision to conduct semi-structured interviews was done because of their capacity to unravel insights with regards to how people perceive the object of study (Bryman, 2012), in this case, to how train-cyclists perceive route decision making according to bike-train route-related attributes. These interviews offered a compelling potential for being attuned to the uniqueness of different route scenarios.

31_{This idea was inspired from reading a recent book written by David Bissell (2018), a thorough qualitative study which explores people’s} everyday transit commuting experiences. Likewise, this book in general gave me the ideas and inspiration to draft the findings for this research sub-question.

(24)

24

Each interview lasted approximately one hour. Once they were all carried out, a verbatim transcription32

was done for the five of them by capturing the exact wording and reactions of participants including for example laughter, pauses, and moments of silence. Thereafter, a line-by-line analysis of the interview transcripts was conducted based on the codes outlined in the operational scheme (see Table 6). Different quotes were gathered from the interviews which were then assembled according to these codes. As Charmaz (2006:50) argues, early coding of interview data helps to “gain a closer look at what participants say and, likely, struggle with.” In turn, this method was chosen because it allowed to refocus the interviews as they were being conducted. In the end, interpretations and conclusions were drawn on the basis of all coded data.

Table 6: The operational scheme for this study.

32_{All participants granted their permission to record the interviews, transcribe them, and to use quotes from their narratives by} signing an informed consent letter (see Appendix F).

(25)

25

4. Bicycle-train system performance

This chapter will portray the system performance of the bike-train mode by discussing the following research sub-question: How does the current bike-train system perform? The succeeding sections represent the different stages that were undergone to answer this question. Section 4.1 briefly explains the data processing that was realized to empirically assess the question. Section 4.2 exhibit the GIS analyses that were carried out for the NS railway stations located within the AMA. Section 4.3 displays three hypothetical travel situations of train-cyclists, supplemented by route maps per each trip segment. Lastly, Section 4.4 discusses the general findings.

4.1 Preparing the data

All the data was collected in a shapefile format (.shp) which is mainly used for ArcGIS desktop applications. It was then exported to a geodatabase33_{as feature classes}34_{in order to start processing the information.}

First, the data of all Dutch municipalities was filtered out to obtain only the municipalities that form part of the Amsterdam Metropolitan Area (AMA). This was done to have a polygon feature of the AMA as the study area which was used for the analyses conducted. Afterwards, the data with all the NS railway stations was also filtered because it included both metro and train stations as well as tram stops. A query expression was done using the ‘Select by attributes’ tool to identify only the NS railway stations. Consequently, using the geoprocessing tool called ‘Clip’, the AMA was used as a base to snip only the stations within its confines.

4.2 Mapping the bike-train system

Density analysis

The AMA has 52 NS train stations which are relatively dispersed. In order to illustrate how some of them are clustered across space, the content management system known as ArcGIS Online was used to perform a density analysis. This type of analysis takes a certain amount of point or line features and spreads their density in a given area.

Firstly, the data of the NS train stations had to be uploaded from ArcMap to ArcGIS Online. Subsequently, their density was calculated using the so-called ‘Calculate density’35_{tool which, by default, calculated in}

square kilometers an appropriate search distance for all stations and classified the output data with 10 classes. The density was symbolized based on different shades of purple in which the darker areas represent the places where most railway stations are agglomerated (See Figure 5). The densest area was

33_{http://desktop.arcgis.com/en/arcmap/10.3/manage-data/geodatabases/what-is-a-geodatabase.htm} 34_{http://desktop.arcgis.com/en/arcmap/10.3/manage-data/geodatabases/feature-class-basics.htm} 35_{http://doc.arcgis.com/en/arcgis-online/analyze/calculate-density.htm}

(26)

26

found in the center of the AMA between the municipal borders of Amsterdam, Diemen, and Ouder-Amstel. The second densest area was located at the west of the AMA between the municipal borders of Haarlem, Velsen, and Bloemendaal.

Figure 5: Density of NS railway stations within the AMA. Source: Author’s illustration.

Overlay analysis

Train stations’ catchment area are contingent upon the mode of transport that travelers use to access them. For an average pedestrian, walking distances are generally understood to be around 0.5 through 1 km. On the other hand, cycling distances are considered to be around 5 to 7 km which by implication increases the spatial reach of stations. In order to compare both catchment areas, their level of increase was verified in terms of impact areas that the stations had in the AMA which depended on whether a person walks or cycles to/from a train station. Consequently, a buffering method was conducted targeting the 52 NS stations (See Figure 6). This was done using the geoprocessing tool called ‘Buffer’ which created two polygon features: (1) a walking catchment area of 1 km and (2) a cycling catchment area of 5 km for all the stations. Afterward, these catchment areas were overlaid on top of the whole AMA region to calculate how much area did each of them covered. When the cycling distance was estimated,

(27)

27

approximately 70.1% of the whole area of the AMA (1793 km²) was identified as the bike-train catchment area (See Table 7). This was around 8 times bigger than the walking-based catchment area.

Access distances Walking (1 km) Cycling (5 km)

Station impact area km²

(% out of the whole AMA area) 154 (8.5%) 1258 (70.1%) Table 7: Comparison of station impact area between walking and cycling distance in the AMA.

Figure 6: NS stations impact area by access distances in the AMA. Source: Author’s illustration.

Proximity analysis

The findings provided by the previous analyses already suggests how the AMA offers a great amount of NS stations of which train-cyclists can choose from when making their route choices. To illustrate how stations’ cycling-based catchment area are overlapping with each other, a buffering method was also conducted by determining a 5 km distance for all NS stations within the AMA. This was done using ArcGIS Online instead of ArcGIS for Desktop because the former provides an output feature that exhibits overlapping areas in different shades of a color. The map below (Figure 7) exhibit results which are closely related to the above density map. However, in this map it can be seen how there are two areas that have

(28)

28

a greater amount of train stations in terms of proximity, one at the west with 8 stations and the other one at the center of the AMA with 12 stations. These overlapping catchment areas are what characterize the system performance of the bike-train mode in terms of its competence for providing opportunities for mobility.

Figure 7: Overlap of NS railway stations’ cycling-based catchment area. Source: Author’s illustration.

4.3 Hypothetical travel situations

Bike-train route choices will vary depending on train-cyclists’ living location and on their travel purpose. In order to understand how they navigate in accordance to the system performance of the bike-train system, three hypothetical travel situations were predetermined which illustrate the different route options that are available when getting from A to B. Only cycling routes equal or less than 5km were mapped whereas for the train routes only the ones with the shortest travel time were considered. The home-end and activity-end route maps below exhibit different routes as indicated by Google Maps’ app, illustrating in color red the fastest routes. Furthermore, each travel scenario has two tables, the first one showing the route options available per trip segment and the second one showing the total combination of bike-train route choices. These scenarios were differentiated using a simple/complex relation based on the amount of overlapping railway stations that each train-cyclist had in close proximity.

(29)

29

1

st

_{scenario: simple-simple}

Karl, a man who lives in Boswijk36_{, needs to make a bike-train trip to the city of Weesp in order to work.}

From his home he only has the option to reach the intercity station called Lelystad Centrum and can pick between two cycling routes that take him to the station. These two available bike routes overlap in some segments which can be seen in Figure 8. Consequently, for the train trip, Karl has only one train station to egress from in proximity to his workplace which is Weesp, the only local railway station at the municipality (Figure 9). The fastest way from Lelystad Centrum to Weesp takes him 37 minutes via an intercity train which then has one transfer to a sprinter train at Almere Centrum. Once in Weesp, he is also able to choose between two different bike routes (Figure 10). In total, Karl has four bike-train routes to choose from which have almost the same travel times (Table 9).

Trip side Routes Travel time (minutes)

Home-end trip Route #1 5

Route #2 5

Train trip Lelystad Centrum – Weesp #1

(1 transfer) 37

Activity-end trip Route #1 6

Route #2 7

Table 8: Bike-train route choices per trip segment for traveler #1.

Bike-train routes choices Total travel time

1 Home-end route #1 + Lelystad Centrum-Weesp + Activity-end route #1 48

2 Home-end route #1 + Lelystad Centrum-Weesp + Activity-end route #2 49

3 Home-end route #2 + Lelystad Centrum-Weesp + Activity-end route #1 48

4 Home-end route #2 + Lelystad Centrum-Weesp + Activity-end route #2 49 Table 9: Total combination of bike-train route choices for traveler #1.

(30)

30

Figure 10: Activity-end routes for traveler #1. Source: Author’s illustration.

Figure 9: Train route for traveler #1. Source: Author’s illustration.

Figure 8: Home-end routes for traveler #1. Source: Author’s illustration.

(31)

31

2

nd

_{scenario: simple-complex}

Aqua, a woman who lives in Toolenburg, Hoofddorp37_{will make a bike-train journey to Amsterdam in}

order to meet with a friend and go to the zoo called Natura Artis Magistra (ARTIS). From her home she only has in close proximity the local railway station called Hoofddorp. Aqua can pick between three bicycle routes in direction towards this station which are not too different in travel time and are also overlapping in some segments (Figure 11). From Hoofddorp station she can choose to egress in four train stations that are near to the zoo (Figure 12). These train routes vary in travel time and some of them have transfers while only one has a direct connection. Likewise, from all the egress stations she can pick between 11 activity-end routes which have different travel times. Some of these routes are overlapping in some portions as it can be seen in Figure 13. In total, Sarah is able to choose between 33 different bike-train routes from her home to ARTIS (Table 11).

Trip side Routes

Travel time (minutes) Home-end trip Route #1 11 Route #2 12 Route #3 14 Train trip

Hoofddorp - Amsterdam Central Station

(direct) 23

Hoofddorp - Amsterdam Science Park

(1 transfer) 36

Hoofddorp - Amsterdam Amstel

(2 transfers) 37

Hoofddorp - Amsterdam Muiderpoort

(1 transfer) 33

Activity-end trip

Amsterdam Central Station - ARTIS

Route #1 8

Route #2 11

Route #3 11

Science Park - ARTIS Route #1 16

Route #2 16

Amsterdam Amstel - ARTIS

Route #1 12 Route #2 12 Route #3 12 Muiderpoort - ARTIS Route #1 8 Route #2 8 Route #3 11

Table 10: Bike-train route choices per trip segment for traveler #2.

(32)

32

Bike-train routes choices

Total travel time

1 Home-end route #1 + Hoofddorp-Central Station + Activity-end route #1 42

10 Home-end route #1 + Hoofddorp-Science Park + Activity-end route #1 63

16 Home-end route #1 + Hoofddorp-Amstel + Activity-end route #1 60

22 Home-end route #3 + Hoofddorp-Amstel + Activity-end route #1 63

25 Home-end route #1 + Hoofddorp-Muiderpoort + Activity-end route #1 52

(33)

33

Figure 13: Activity-end routes for traveler #2. Source: Author’s illustration.

Figure 12: Train routes for traveler #2. Source: Author’s illustration.

Figure 11: Home-end route for traveler #2. Source: Author’s illustration.

(34)

34

3

rd

_{scenario: complex-complex}

Alexa, a student who lives in Planetenwijk, Haarlem, needs to make a bike-train trip to Amsterdam because she has a lecture in Roeterseiland Campus at the University of Amsterdam (UvA). Since her home is located in an overlapping catchment area, she has the option to choose between six stations which are in close proximity. To reach these stations, there are a total of 17 home-end routes which vary in travel time and most of them are overlapping with each other (Figure 14). From all the access stations, Alexa can also pick between six egress stations that are near to the UvA (Figure 15). These train routes have different travel times, some have transfers while others are direct connections. Likewise, once Alexa arrives to any of the destination stations, she has a total of 18 activity-end routes to choose from which are also overlapping in some segments (Figure 16). In total, she is able to make 304 bike-train route combinations from her home in Haarlem to the UvA. Table 13 shows only the fastest routes; Appendix G exhibits the total amount of bike-train route choices for this scenario.

Trip side Routes

Travel time (minutes) Home-end trip Haarlem Route #1 7 Route #2 10 Bloemendaal Route #1 6 Route #2 7 Route #3 8 Overveen Route #1 15 Route #2 15 Route #3 19 Santpoort-Noord Route #1 14 Route #2 15 Route #3 20 Santpoort-Zuid Route #1 10 Route #2 11 Route #3 12 Haarlem Spaarnwoude Route #1 15 Route #2 16 Route #3 16 Haarlem Amsterdam Zuid (2 transfers) 42 Amsterdam Muiderpoort (1 transfer) 29 RAI (2 transfers) 48

Amsterdam Science Park

(1 transfer) 32 Amsterdam Central Station (direct) 15 Amsterdam Amstel (1 transfer) 27 Amsterdam Zuid (2 transfers) 60

How do train-cyclists navigate? : exploring bike-train route choice behavior in the Amsterdam Metropolitan Area

= ?

Author:

Pedro Nieves

Student number:

11659912

Email:

pgna.pr@gmail.com

Submission Date:

11 June 2018

Supervisor:

Dr. Marco te Brömmelstroet

Second reader:

Dr. Pieter Tordoir

Master’s program:

Urban & Regional Planning

How do train-cyclists

navigate

?

Exploring bike-train route choice behavior in the

Amsterdam Metropolitan Area

University of Amsterdam

Faculty of Social and Behavioral Sciences

Graduate School of Social Sciences

How do train-cyclists navigate?

Exploring bike-train route choices in the

Amsterdam Metropolitan Area

Supervised Research Project

Written by Pedro Nieves

In partial fulfillment of the graduation requirements for the

Master of Science in Urban and Regional Planning

Acknowledgments

First and foremost, I would like to thank my supervisor, Marco te Brömmelstroet (a.k.a. the

fietsprofessor), for his guidance, inspiration, and constant feedback throughout my research. I

am extremely grateful for his availability, attentiveness, and for his constructive criticism which

challenged me to think outside of the box, encouraging me to write a better thesis within the

timeframe.

I would also like to thank my colleagues from the University of Amsterdam with which I had the

chance to discuss my study. I really appreciated their personal support and remarks throughout

these past six months. Their friendship, cheerfulness, and dedication have inspired me and made

this an unforgettable experience.

The past months living in Amsterdam have been very challenging yet rewarding – I came out of

this program as a different person, feeling inspired and ready to contribute towards a better

future in my home(is)land, Puerto Rico.

Finally, I want to express my deepest gratitude to my friends and family back home for their

unconditional love, support, and comfort during my studies. Special thanks to my sister,

Alexandra, and to my parents who have always been there for me throughout my academic

career. I share this achievement with you. Gracias por tanto.

Sincerely,

_______________________

Pedro Nieves

Abstract

Resumen

Table of Contents

1.

Introduction

2.

Theoretical Framework

2.1 Bike-train system

2.2 Route choice behavior

2.2.1 Bicycle route choice

2.2.2 Transit route choice

2.3 Conceptual framework (preliminary)

3.

Methodology

3.1 Problem Statement

3.2 Research Question(s)

3.3 Study Area

3.4 Research Design

3.4.1 How does the current bike-train system perform?

3.4.2 What are the (in)direct attributes which influence potential bike-train route choices?

3.4.3 To what extent are train-cyclists willing to trade-off travel time with other bike-train route

choice attributes?

3.4.4 How are bike-train route-related attributes playing a role in train-cyclists’ route choice

behavior?

4.

Bicycle-train system performance

4.1 Preparing the data

4.2 Mapping the bike-train system

Density analysis

_?

_{scenario: simple-simple}

_{scenario: simple-complex}

_{scenario: complex-complex}