Urban structural and socioeconomic effects on local, national and international travel patterns and greenhouse gas emissions of young adults

Journal of Transport Geography 68 (2018) 130–141

Contents lists available at ScienceDirect

Journal of Transport Geography journal homepage: www.elsevier.com/locate/jtrangeo

Urban structural and socioeconomic effects on local, national and international travel patterns and greenhouse gas emissions of young adults

T

⁎

Michał Czepkiewicza,c, , Juudit Ottelina,b, Sanna Ala-Mantilaa,b, Jukka Heinonena, Kamyar Hasanzadehb, Marketta Kyttäb a

Faculty of Civil and Environmental Engineering, University of Iceland, Reykjavik, Iceland Department of Built Environment, Aalto University, Finland c Institute of Geoecology and Geoinformation, Adam Mickiewicz University in Poznań, Poznań, Poland b

A R T I C LE I N FO

A B S T R A C T

Keywords: Long-distance travel Greenhouse gas emissions Travel behavior Urban structure Carbon footprint

The inverse relationship between urban density and greenhouse gas (GHG) emissions caused by driving is well established. However, at the same time the few existing studies have observed higher levels of long-distance travel and particularly air travel in the same densely built parts of urban regions. This may lead to GHG emissions reduction in local travel offset by the concomitant increase in long-distance travel. With this study we aim to identify the main factors involved in differences in local, national and long-distance travel patterns and the resulting GHG emissions, with a special focus on the role of the different urban zones in the Helsinki Metropolitan Area (HMA) in Finland. We used a softGIS survey to collect data on the personal travel of young adults living in HMA. SoftGIS methodology provides the opportunity to obtain detailed spatial data on participants' residential locations, travel destinations, and destination characteristics such as travel modes, frequencies and trip purposes. Special attention was paid to national and international trips, for which data were collected over 12 months, a period long enough to capture actual travel patterns. GHG emissions were assessed with a wide scope life cycle assessment (LCA) approach, including vehicles and infrastructure, and the results were elaborated with a two-part regression model on participation in travel and amount of GHG emissions. The study found that the residential location was associated with travel emissions on all scales, and independently from major socioeconomic characteristics. Residents of centrally located and densely built urban zones have on average lower emissions from local travel but higher emissions from international travel than residents of caroriented suburban zones, and the association holds true after controlling for income, education level and household type. Differences in emissions from local travel between most central and most suburban zones were almost completely offset by differences in emissions from international travel. International long-distance trips were a dominant source of travel-related GHG emissions in all urban zones, particularly due to plane flights.

1. Introduction Energy usage and greenhouse gas (GHG) emissions caused by personal travel in various urban settings have been broadly studied (e.g. Ewing and Cervero, 2010; Newman and Kenworthy, 1999; Mindali et al., 2004; Norman et al., 2006; Chapman, 2007; Brownstone and Golob, 2009; Brand and Preston, 2010; Chester et al., 2013). While the inverse relationship between urban density and GHG emissions caused by road traffic is well established (e.g. Næss, 2012; Ewing and Cervero, 2010; Newman and Kenworthy, 1989, 1999; Mindali et al., 2004; AlaMantila et al., 2016), the relationship between urban form and longdistance travel has been much less studied. However, several studies so far have observed higher levels of long-distance travel, and particularly

⁎

air travel, in big cities and densely built parts of urban regions (Brand and Preston, 2010; Holz-Rau et al., 2014; Ottelin et al., 2014; Reichert et al., 2016). Previous studies have suggested various hypotheses for the reasons behind this relationship. Frändberg and Vilhelmson (2011) attributed the increase in international travel and decrease in car-dependency among Swedish young adults to globalization and flexibilization of lifestyles. Large and globally-connected cities, with good educational and career opportunities, are especially attractive to highly skilled individuals, who not only have resources for travelling but also may be more cosmopolitan in their lifestyles (Reichert and Holz-Rau, 2015; Reichert et al., 2016). Another reason for the higher amount of long-distance travel among big city residents may be better access to airports, as illustrated by Bruderer Enzler (2017). However, these

Corresponding author at: Faculty of Civil and Environmental Engineering, University of Iceland, VR-II, Hjarðarhaga 6, 107 Reykjavík, Iceland. E-mail address: [email protected] (M. Czepkiewicz).

https://doi.org/10.1016/j.jtrangeo.2018.02.008 Received 12 October 2016; Received in revised form 23 January 2018; Accepted 20 February 2018 0966-6923/ © 2018 Elsevier Ltd. All rights reserved.

Journal of Transport Geography 68 (2018) 130–141

M. Czepkiewicz et al.

in GHG emissions between inhabitants of the urban zones were solely due to their sociodemographic, economic and life situation characteristics, or there was also a separate effect of urban structural variables. We also discuss how the findings related to the compensation hypothesis, rebound effects, and cosmopolitan lifestyles, all of which have been suggested as potential reasons for such differences in travel behavior, but were not systematically tested in this study.

explanations are more likely to apply to urban-rural differences than to differences between various urban zones, which is the focus of this study. Ottelin et al. (2014) and Heinonen et al. (2013a,b) have suggested a monetary-based rebound as another possible explanation that is more closely connected to urban form. Since car ownership and operation are quite expensive, people who do not have these expenses may have higher expenditures on other goods and services, especially holiday travel (see also Lenzen et al., 2004; Ornetzeder et al., 2008; Chitnis et al., 2013; Heinonen et al., 2013a, 2013b). Some researchers have also discussed the compensation hypothesis: the residents of densely populated urban areas may tend to compensate for the lack of open space, green areas, and recreational opportunities by taking longer and more distant holiday or weekend trips (Holden and Norland, 2005; Strandell and Hall, 2015). The purpose of the study was to closely examine local, national, and international travel patterns of young adults who differ in sociodemographic and life situation characteristics and live in different urban zones of Helsinki Metropolitan Area (HMA). The study extended the research of [citations concealed] about GHG emissions from transport activities of urban dwellers in Finland by enabling more comprehensive data collection on travel purposes, modes and destinations. The previous census data-based studies have suggested substitution between car ownership, local travel and long-distance travel. However, these studies suffered from data gaps, such as short-term reference periods in travel diary surveys and the lack of detailed spatial information on participants and their trips (Ottelin et al., 2014; HolzRau et al., 2014; Reichert et al., 2016). In the current study we used a softGIS survey to collect data on local, national and international travel of young adults (25- to 40-yearolds) living in the Helsinki, Espoo, Vantaa and Kauniainen municipalities within the Helsinki Metropolitan Area (HMA). Geographical data on mobility patterns and environmental evaluations were contributed by the respondents themselves through an online questionnaire coupled with an interactive map (Kahila and Kyttä, 2009; Kyttä, 2011; Brown and Kyttä, 2014). SoftGIS methodology has been used previously in mobility-related studies (e.g. Salonen et al., 2014; Haybatollahi et al., 2015). The survey method allowed us to obtain detailed spatial data on participants' residential locations and trip destinations, and helped to avoid some of the caveats of previous mobility studies related to spatial unit aggregation (Kwan, 2012). It also allowed us to ask contextual questions related to the trips, such as their purpose, frequency, and travel mode. Our research covered a 12-month period for national and international trips, which was enough to overcome the shortcomings of some previously used datasets (Ottelin et al., 2014). The GHG emissions assessment was conducted with the life cycle assessment (LCA) approach, including both direct and indirect emissions. Inclusion of the indirect or upstream emissions, up to the production of the utilized vehicles and construction of the transport infrastructure, is an important feature adding value to this study, since these have been found significant in the context of transport (Chester and Horvath, 2009; Chester et al., 2013), but have been excluded from most of existing studies. Furthermore, the employed GHG emissions factors included both the typical long-lived GHGs (LLGHGs) included in standard LCAs, as well as the short-lived climate forcers (SLCFs) for air travel, which account for an important part of the climate impact of aviation (Lee et al., 2009; Borken-Kleefeld et al., 2013; Peters et al., 2011; Aamaas et al., 2013; Aamaas and Peters, 2017).

1.2. Scope of analysis The study was limited to young adults (25- to 40-year-olds) living within the HMA. The choice of the age group was meant to minimize the effect of life course variables on the analysis, as its members predominantly belonged to an active workforce, were already independent adults, but were still well before retirement age. Furthermore, people of this age have grown up in a globalized world with access to information and communication technologies, which may explain differences in their travel patterns compared to those of older adults. Finally, understanding the travel patterns of younger people is especially relevant for the design of future sustainable built environments. The analysis of national and international travel patterns focused on trips related to leisure and family visits. Business trips were not included, because they normally constitute a relatively small share of long-distance travel, and are influenced by factors other than those affecting leisure trips (Reichert et al., 2016; Bruderer Enzler, 2017). Analysis of local travel included both leisure and work- or study-related travel. The paper is structured as follows: the next section presents the Research materials and methods, followed by the Results section. The paper ends with a Discussion and conclusions, including Future Research directions. 2. Research materials and methods 2.1. Sampling and data collection The data were collected using softGIS, a public participation GIS (PPGIS) method that combines online questionnaires with interactive maps (Brown and Kyttä, 2014). In this study, we used softGIS methodology to record travel patterns. Study participants marked destinations of local, national, and international trips as points marked on an interactive map. Each marked location was accompanied by questions on trip characteristics such as purpose, travel mode or frequency of visits (SI Figs. 1 to 3). The survey also included questions not related to a map, such as background information about respondents and their household, car ownership and characteristics (fuel efficiency and annual mileage), residential location and dwelling characteristics, satisfaction with life domains, pro-environmental behaviors, consumption patterns, and personal attitudes. The questionnaire that we used for this study consisted of 12 thematic pages and was distributed in two language versions: Finnish and English. Only background information and travel patterns were used in this study. The target population of the survey were all inhabitants of the HMA (consisting of municipalities of Helsinki, Vantaa, Espoo, and Kauniainen) aged from 25 to 40 years. The Population Register Center of Finland provided addresses for a random sample of a target population living in the region. A total of 5000 individuals were contacted by letter in August 2016 and asked to use the Internet to answer the questionnaire. A second round of invitations was sent out later during September of the same year. Altogether 962 respondents replied to the questionnaire. Because of incomplete responses, we limited the sample used in the analyses to 841 respondents (response rate 16.82%).

1.1. Research goals and questions In line with the overall purpose of the study, the first specific goal was to compare the GHG emissions related to local, national and international travel by young adults who inhabit different urban zones of the HMA. The second, and the main, goal was to identify the key factors in the differences in travel-related GHG emissions among the studied population. We sought to provide evidence on whether the differences

2.2. Study area and urban zones HMA is the capital region of Finland and the most populous 131

Journal of Transport Geography 68 (2018) 130–141

M. Czepkiewicz et al.

Fig. 1. Urban zone classification of Helsinki metropolitan region based on an open Travel-related Urban Zones data set by the Finnish Environment Institute SYKE (2010).

bus stops and 400 m for rail stops. All the cells that have > 60 min average stop frequency or > 250 m for bus stops and > 400 m for rail stops are classified as car zones (Ristimäki et al., 2011, 2013). We assigned each of the survey respondents to an urban zone based on the residential location provided in the questionnaire. The respondents inhabiting the urban zones differed in their main background variables (Table 1). The car zone had the highest share of highest income respondents (27%), whereas pedestrian zones of the sub-centers had the lowest share of this group (13%). Furthermore, the highest education levels were observed in the most central zones. Car ownership patterns were as expected: about 37% of the respondents had a car in the household in the central pedestrian zone and its fringe, whereas in the car zone the share of car owning households were 79%. Suburban zones also host a higher percentage of families with children and a lower percentage of singles and childless couples than the central zones.

metropolitan region with approximately 1.1 million inhabitants (OSF, 2016). Even though rather small if compared internationally, it is a well-connected global city, and an important economic and cultural center. More than one third of Finland's GDP is produced in the region, and it is the location of the main airport of the country. The urban region is characterized by a strong commercial and cultural role of its urban core. The region also has local sub-centers and public transportation corridors (see Fig. 1). The urban core is densely built, but the outer areas are typically more sprawling and car-dependent. A comprehensive public transportation network is in place, thus limiting the necessity to use private cars for local travel. Compared to many other metropolitan areas around the world, HMA is characterized by a relatively affluent urban core, and a mixed socioeconomic structure of the suburbs. Simply put, those in the west are typically wealthier, whereas many areas in the east are less affluent. However, compared globally, socioeconomic segregation in the area is low. To differentiate between travel-relevant urban structure types of the region, we used the Travel-related Urban Zone classification provided by the Finnish Environment Institute SYKE (2010). The GIS-based (250 × 250 m grid of cells) classification divides urban regions into zones according to their location relative to the city center, population characteristics, public transport supply, building stock and employment (for more information see Söderström et al., 2015). The classification includes nine zones, aggregated in this study into six due to respondents' distribution and better result interpretability: the central pedestrian zone, the fringe of the pedestrian zone, the pedestrian zones of the sub-centers, intensive public transport zone, basic public transport zone, and the car zone (Fig. 1). The central pedestrian zone and its fringe are characterized by short distances and good walking access to services and jobs, is densely built and with a mixed urban structure, and has good conditions for both non-motorized and public transportation. The pedestrian zones in the sub-centers are located around local centers and characterized by very good public transportation access, a high number of jobs in retail and a high population density. Intensive and basic public transportation zones and the car zone are differentiated based on the level of public transportation supply. In the intensive zone, the frequency of stops is on average 5 min for buses and 10 min for rail transit, and the distance to the nearest stop is a maximum of 250 m for

2.3. Destinations and travel distances Respondents were asked to provide details of their travel patterns on four separate pages of the questionnaire. On page 4, they provided the characteristics and locations of their residences and main work or study places. Page 5 consisted of questions related to location and characteristics of places visited within HMA (i.e. local trips). The respondents were asked to mark between 5 and 15 locations they had been frequently visiting. The time frame was not specified to capture habitual travel patterns. Participants marked visited locations in seven categories: work or study places, services and errands; shopping; leisure and going out; culture and sport events; daycare, kindergarten or school; sports and active recreation. Each marked location was associated with questions about travel mode, frequency of visiting, and direction of travel (i.e. whether it was visited from home, work or study place, or on the way between home and work/study place). Pages 6 and 7 consisted of questions related to destinations visited within Finland but away from the HMA (i.e. national trips) and destinations visited abroad (i.e. international trips). On both pages, the participants were asked to mark all trips made during the 12 months prior to the data collection. Each marked location was associated with questions regarding number of trips made to the location, trip purposes, 132

Journal of Transport Geography 68 (2018) 130–141

M. Czepkiewicz et al.

Table 1 Means and proportions of background variables in HMA urban zones. Urban zone

Central pedestrian zone

Fringe of central pedestrian zone

Pedestrian zones of sub-centers

Intensive public transportation zone

Basic public transportation zone

Car zone

Total sample size Na Income < 1500 EUR 1500–3000 EUR 3000–4500 EUR 4500–6500 EUR > 6500 EUR Education Basic and secondary Lower tertiary Graduate and post-graduate Household type Single Couple living together Families with children Other Car ownership % Number of cars (if car in the household)

93

182

102

163

182

119

12% 31% 21% 16% 20%

11% 29% 18% 24% 18%

9% 30% 23% 25% 13%

12% 21% 29% 18% 20%

8% 18% 24% 27% 23%

5% 19% 21% 29% 27%

12% 31% 57%

14% 36% 50%

22% 38% 40%

22% 38% 40%

23% 36% 41%

22% 42% 36%

42% 39% 18% 1% 37% 1,13

35% 42% 20% 3% 38% 1,07

27% 36% 34% 4% 64% 1,13

32% 34% 31% 2% 55% 1,20

22% 28% 49% 2% 72% 1,24

18% 28% 50% 4% 79% 1,31

a

Sample sizes for each variable are not identical to total sample size N; pair-wise deletion was used in case of missing values.

2.4. Estimation of greenhouse gas emissions

and number of interchanges for plane trips. National trips were grouped in four categories by travel mode: car, train, bus, and plane. International trips were grouped in five categories: car, train, bus, boat and plane. Distances between homes and destinations visited by boats and planes were calculated as geodesic shortest distances in a Spatialite database using the World Geodetic System 1984 (WGS84) coordinates to consider the curvature of the Earth. Distances to destinations visited by car, bus, train, foot and bicycle, were calculated along road and rail infrastructure networks. For local trips we used Digiroad, a road dataset provided by the Finnish Transport Agency, which includes all motorway and pedestrian routes in Finland, but we excluded the nonmotorized routes to include only driving distances. As a result, distances covered on foot or by bicycle may be overestimated, but it does not affect the calculation of emissions, as these travel modes generate none. For national and international trips by car and bus we used the Global Roads Open Access Data Set compilation provided by the Center for International Earth Science Information Network (2013). For national and international trips by train we used a railroad network layer of Vector Map Level 0. We then calculated distances between the destinations and residential locations marked by respondents, using the shortest path algorithm in the Network Analyst toolbox in ArcGIS Desktop 10. For national and international destinations visited by car or bus, for which calculating network distances was not possible (e.g. located on a different continent or too far away from road network), we used the geodesic shortest distances. Local trip frequencies were measured in categories related to weekly or monthly periods (e.g. “five to seven times a week” or “once or twice a month”) and coded numerically to estimate the number of trips made during 12 months. The reported number of national and international trips were also coded numerically and used to estimate the number of trips in 12 months. Every national and international destination was treated as a two-way trip originating in the HMA. The reported numbers of flight interchanges were used to assign flights to distance categories and to correct distance estimation by multiplying by 1.2 for each interchange. Estimated yearly distance was then calculated for each of the marked destinations. The yearly distances were then multiplied by GHG emission coefficients described below, according to travel mode. We discuss uncertainties related to distance calculation in Section 4.1.

The GHG assessment was conducted with a life cycle assessment (LCA) approach, meaning that both the direct and the indirect emissions were included. With the indirect emissions, a wide scope was adopted, including the GHG emissions from producing the fuel and electricity (for electric vehicles) used, but also vehicle manufacturing and infrastructure construction, in accordance with to the suggestion of Chester and Horvath (2009), who have found them as major contributors to the GHG emissions from transport. The GWP100 metric was utilized, but in addition to the LLGHGs typically included in GWP calculations (following the Kyoto Protocol), we included the short lived climate forcers (SLCFs) to the air travel assessments, which are highly important in aviation as shown by, e.g., Aamaas et al. (2013), Peters et al. (2011) and Borken-Kleefeld et al. (2013). For ground transport, the GWP100 factors are dominated by the LLGHGs and omission of SLCFs does not alter the assessment outcomes (Peters et al., 2011). Three main emission data sources were utilized. The direct combustion phase emissions were taken from the LIPASTO database produced by the VTT Technical Research Centre of Finland Ltd. for ground transport for different vehicle types used in Finland (VTT, 2016). For air travel, the combustion phase emissions were taken from Aamaas et al. (2013), and the split into short (< 800 km) and long (> 800 km) flights follow the source. The values are considerably higher than values without SLCFs provided by VTT (2016), where emissions are estimated at 0.26 CO2e kg/PKT for flights shorter than 463 km, and at 0.11 CO2e kg/PKT for flights above 3000 km. Therefore, inclusion of SLCFs emphasizes the importance of emissions caused by air travel, and longhaul flights in particular. Furthermore, the indirect emissions were added, following Chester and Horvath (2009). They include roadways, tracks, stations, runways and other infrastructure, vehicle production and maintenance and fuel production. The uncertainty imposed by the assumption that the indirect emissions are compatible between the U.S. and Finland is discussed in the uncertainty analysis. When necessary, an equivalent for the average occupancy rate (see Chester and Horvath, 2009) in Finland was calculated. Furthermore, as Chester and Horvath (2009) do not report indirect emissions for ferries, we have used the values for midsize aircraft due to relative similarities in infrastructures for these travel modes, and lack of other sources. For trips with private cars, the fuel efficiencies reported by the survey respondents for the primary car of the household were used for the assessment. The fuel efficiency was asked with a five-category 133

Bus

Local trips

National and international trips

Car

All trips

134 0.240 kg/PKT

> 800 km

Ferry

0.300 kg/PKT

< 800 km

Plane

0.223 kg/PKT 0.400 kg/PKT

Helsinki-Stockholm, average occupancy Helsinki-Tallinn, average occupancy

0.022 kg/PKT

Train

0.049 kg/PKT

0.011 kg/PKT

0.054 kg/PKT

0.022 kg/PKT

2.36 kg/l 1.3 occupancy for local trips 1.9 for all other trips 0.069 kg/PKT

Direct emissions (Combustion)

Diesel bus, average occupancy rate on long distance trips, 12/50 passengers Pendolino and intercity trains, average occupancy

Average occupancy rate in local traffic

Metro

Bus

Average occupancy rate in local traffic

Natural gas bus, average occupancy rate in local traffic, 18/50 passengers Average occupancy rate in local traffic

Reported fuel efficiency for 1st vehicle

Explanation

Tram

Commuter train

Travel mode

Travel scope

Table 2 GHG emission coefficients per travel mode and sources used in calculations (all units in CO2e).

VTT (2016)

VTT (2016)

Aamaas et al. (2013)

Aamaas et al. (2013)

VTT (2016)

VTT (2016)

VTT (2012)

VTT (2012)

VTT (2012)

VTT (2012)

U.S. EPA (2008)/VTT (2016)

Source

0.015 kg/PKT

Included in combustion factor Included in combustion factor Included in combustion factor 0.015 kg/PKT

Included in combustion factor Included in combustion factor Included in combustion factor 0.037 kg/PKT

0.031 kg/PKT

0.026 kg/PKT

Fuel production

Indirect emissions

0.020 kg/PKT

0.020 kg/PKT

0.020 kg/PKT

0.020 kg/PKT

0.062 kg/PKT

0.058 kg/PKT

0.083 kg/PKT

0.094 kg/PKT

0.062 kg/PKT

0.050 kg/PKT

0.074 kg/PKT

Life cycle

Chester and Horvath (SFBA Caltrain) Chester and Horvath (midsize aircraft) Chester and Horvath (midsize aircraft) Chester and Horvath (midsize aircraft) Chester and Horvath (midsize aircraft)

Chester and Horvath (SFBA Caltrain) Chester and Horvath (Boston Green Line) Chester and Horvath Muni) Chester and Horvath

(2009)

(2009)

(2009)

(2009)

(2009)

(2009)

(2009) (SF

(2009)

(2009)

Chester and Horvath (2009)

Chester and Horvath (2009)

Source

M. Czepkiewicz et al.

Journal of Transport Geography 68 (2018) 130–141

Journal of Transport Geography 68 (2018) 130–141

M. Czepkiewicz et al.

Fig. 2. Estimated annual GHG emissions from local, national and international travel of young adults living in different urban zones of Helsinki metropolitan area. National and international travel only includes leisure and family-related trips, whereas local travel includes all reported trip purposes.

but not censored at zero. An important question in influencing choice of modeling approach is whether participation in emissions and the amount of emissions come from the same data generating process, i.e. whether the underlying mechanisms are the same. In the case of long-distance travel emissions, the underlying processes may be different. For instance, a decision to travel for holiday may be related to economic factors in a separate way than selection of travel modes or trip destinations. Therefore, we selected a two-part modeling approach, following Min and Agresti (2002) and Fletcher et al. (2005). The first part is a binary model for participation in long-distance travel in which the outcome is dichotomous and modeled by logistic regression. The second part is modeled using an ordinary least squares (OLS) regression on emission values transformed with normal logarithm. We used a stepwise selection of explanatory socio-demographic and economic variables. Our original set of variables included age, gender, household type, education, occupation and income. Occupation and age were dropped from the final regression models, since they correlated with income and education and had low if any explanatory power and statistical significance alone. Also, the limited age frame of the respondents (25 to 40-year-olds) affected the decision to leave out the age groups. The tested urban structure related variables include urban zone, car ownership, type of residence and private yard. We tested the impact of these variables separately, since they correlate with each other and the inclusion of several of these in the same model would make the interpretation of the results ambiguous. We combined the two central urban zones (i.e. the central pedestrian zone and the fringe of the central pedestrian zone) into one class due to their structural similarity, and to achieve stronger statistical significance in the regression models (we tested models with the separate zones first). However, we kept the two zones separated in the descriptive analyses (Figs. 1-3) to illustrate their similarity in the reported travel patterns. In addition to the urban structure related variables, we tested the impact of having business trips and the impact of working hours separately. Our hypothesis was that high amount of business travel could decrease the willingness to travel on one's own time (when income is controlled). The hypothesis related to working hours was that people who work very little would travel less for instance because of the lack of money, but people who work very much, would also travel less than people who work moderately, for instance because of the lack of time or differences in lifestyle orientations.

question with options from below 4 l per 100 km (l/100 km) up to over 10 l/100 km with two-liter intervals and separate options for electric vehicles and “unknown”. In the calculations, below 4 was treated as 3, middle values were taken as 4–6, 6–8 and 8–10 l/100 km, and 12 l/ 100 km was assumed for the category “above 10”. The estimated fuel consumption was turned into GHG emissions with a multiplier of 2.36 kg CO2e/l (US EPA 2008). For the approximately 10% of the sample who did not answer the efficiency question though possessing private cars, the Finnish average GHG intensity factor of 0.167 kg CO2e/km from the LIPASTO database was utilized (VTT, 2016). The Finnish average occupancy rates of 1.3 for local trips and 1.9 for all other trips from the LIPASTO database were employed to turn the vehicle emissions into the per passenger kilometer traveled (PKT) level. Table 2 shows the GHG intensity factors used, the sources and the units for all the travel modes.

2.5. Analysis In the first step of analysis, we performed an exploration of bivariate associations with simple bar charts to identify the main differences in emission levels across sociodemographic and urban structural categories. Reported are also the trip frequencies and average lengths to depict the key differences. The results of these comparisons are presented in subsequent sections. In the second step, we applied regression models to study the impact of socioeconomic and urban structure characteristics on two outcome variables: yearly emissions generated by national leisure or family trips, and yearly emissions generated with international leisure or family trips. The calculations of emissions did not include national and international trips related to work, school or business. However, the impact of participating in such trips on longdistance leisure travel was tested by an explanatory dummy variable. The outcome variables contained multiple cases with no emissions (zero values) and were highly positively skewed. For national travel the number of zeroes was 82 (~10%) and for international travel 170 (~20%). Therefore, the data violate the assumption of normally distributed residuals. Zero emissions in such a dataset signify respondents who did not participate in long-distance travel emissions in the given reference period. In other similar datasets zero emissions may also represent respondents who took part in long-distance travel in each year, but the result was not reflected in data due to a short reference period (e.g. Ottelin et al., 2014; Reichert et al., 2016). Such datasets are censored at zero and require specific analytical methods (Reichert et al., 2016). The data frame in our survey was 12 months prior to the survey, and therefore we treated respondents who did not provide data on longdistance travel as non-travelers. The respondents may also have traveled, but did not mark locations in the survey, for instance because of difficulty in using the mapping tools. To avoid bias, we excluded participants who completed the survey only partially, which yielded 841 records. Thus, we were able to treat the final data set as zero-inflated,

3. Results 3.1. GHG emissions and urban zones GHG emissions for the studied young adults resulting from local travel of suburban residents are much higher than those produced by residents of centrally located and densely built zones (Fig. 2, Table 3). 135

Journal of Transport Geography 68 (2018) 130–141

M. Czepkiewicz et al.

Fig. 3. Modal structure of estimated annual GHG emissions from (left to right) local, national and international travel for young adults in different urban zones of HMA. National and international travel only includes leisure and family-related trips, whereas local travel includes all reported trip purposes.

11 car trips (55% mode share in national trips) per year within Finland compared to 17 car trips (85% mode share) made by the residents of the car zone. Conversely, residents of the central pedestrian zones are more likely to use public transportation in national travel. They make on average 7.4 trips (38% mode share) by train or bus, compared to 2.3 such trips (12% mode share) taken by the residents of the car zone. This may result from differences in car ownership, which is lower in the central pedestrian zone and its fringe than elsewhere (37% compared to 55–79%). However, the differences in car ownership between the zones are not very high (Table 1). The most common travel modes for international trips are plane (52% mode share) and boat (38% mode share). Residents of the central pedestrian zone and its fringe make more international trips per year (5.5 in the central pedestrian zone and 5.8 in its fringe) than residents of other zones (below 5 in all zones, 4.3 in basic public transportation zone). They are also more likely to travel internationally by plane than residents of different zones (ca. 60% of their international trips are taken by plane, whereas the number is close to 48% in suburban zones). Average trip lengths in international travel show no clear patterns, which suggests that destination choices and geographic scope of travel does not vary between the residents of different zones in the same way as trip frequency does, and may be explained by other factors not covered by this study.

Similar patterns have been found in numerous previous studies on daily mobility (Ewing and Cervero, 2010; Næss, 2012). An almost reverse pattern is found in international travel for leisure and family purposes: the highest levels of emissions are generated by the residents of the central pedestrian zone and its fringe (Fig. 2, Table 3). The relative differences between zones in average emissions from international travel are not as high as in local travel. However, due to absolute amounts, lower level of GHG emissions from local travel among residents of central pedestrian zones is almost completely offset by higher levels of GHG emissions from international travel. The overall estimated GHG emissions for the young adults who live in most central zones and most suburban zones are very similar: 4.8 t CO2e/a in the central pedestrian zone and its fringe, and 4.9 t CO2e/a in the car zone. Total travel-related emissions of young adults who live in intensive public transportation zone and pedestrian zones of the sub-centers seem to be lower, at 4.1 t CO2e/a per person. However, a non-normal distribution of emission values and large margin errors (Table 3) do not allow us to conclude about differences in total travel-related emissions between residents of different urban zones. International travel constitutes the largest share of total GHG emissions from travel (46% to 73%) and local trips constitute the smallest share for the residents in most of the zones (5% to 25%). Fig. 3 shows how the emissions from international travel are dominated by flights. The results also illustrate that residential location is not strongly associated with emissions from national travel, except for somewhat higher levels of emissions among residents of the basic public transportation zone (Fig. 2, Table 3). We also analyzed the travel patterns of the studied group in terms of distances, travel modes, and number of trips. Detailed results are provided as supplementary information (SI Tables 1 and 2). Despite minor differences between urban zones in terms of emissions from national travel, there are differences in the use of travel modes. Young adults who live in the central pedestrian zone and its fringe make on average

3.2. GHG emissions and socio-demographic and economic categories Average GHG emissions from travel activities follow also certain economic and socio-demographic characteristics of the respondents. All results are provided in the SI section (SI Tables 3–8), and some key issues are raised here. Income level seems to have, not surprisingly, the strongest effect on the emissions in all the three categories (local, national and international) of travel, increasing continuously with an increase in the monthly income. The pattern is the strongest for car travel

Table 3 Average values of travel-related GHG emissions per year per person (kg CO2e) in HMA urban zones. Urban zone

Central pedestrian zone Fringe of the central pedestrian zone Pedestrian zones of the sub-centers Intensive public transport zone Basic public transport zone Car zone

N

93 182 102 163 182 119

Average travel-related GHG emissions per year per person [kg CO2e] and margin errors ( ± ) at 95% confidence level Local

National

International

Total

233 ± 93 348 ± 76 684 ± 134 807 ± 131 1072 ± 181 1230 ± 192

1057 ± 220 1065 ± 156 903 ± 109 839 ± 65 1425 ± 242 1116 ± 114

3502 3376 2546 2525 2152 2603

4792 4789 4133 4171 4648 4948

136

± ± ± ± ± ±

341 224 259 267 166 219

± ± ± ± ± ±

883 775 1229 575 660 796

Journal of Transport Geography 68 (2018) 130–141

M. Czepkiewicz et al.

and international air travel. In the latter, the annual GHG emissions increase from < 1.5 ton of CO2e per capita in the lowest income group (< 1500 €/month), to approximately 3 tons of CO2e per capita in the most affluent (over 6500 €/month). GHG emissions from air travel still dominate all the emissions for all income groups. Education level has a two-way and less clear effect. The respondents with basic or upper secondary education generate the lowest overall amount of GHG emissions. Emissions from international air travel also follow the education level. Annual GHG emissions of those with basic or secondary education only slightly exceed 1 ton of CO2e per capita while those with graduate and post-graduate degrees on average cause emissions close to 3 tons of CO2e per capita. Working hours seem to both enable travelling and hinder travel activity when high-enough, but the patterns are not very strong. Groups with < 35 weekly working hours have the lowest travel activity in all the three categories in comparison to those working 35–45 a week, with the most significant difference in international travel. However, those working over 45 h a week again cause less GHG emissions than the middle group in all three categories, local, national and international trips, but still more than those working below 35 h a week. Childless households appear to travel the most, especially strongly showing in air travel of households with two adults. In local travel, it is the households with children who cause the highest emissions with their travel activity. People who don't own a car have on average much lower emissions in local travel than people owning one or more cars. In national travel, car ownership is not associated with level of emissions, but it results in different structure: people who don't have cars more frequently use planes and buses at national scale. Differences in structure of emissions from international travel related to car ownership are negligible. Similar pattern is observed for type of residence: people living in apartments had markedly lower emissions from local travel than people living in detached or semi-detached houses, but they have slightly higher emissions from international travel, mostly due to higher level of plane use.

household type and gender have no statistically significant effect on participation in international travel. The likelihood of participating in national travel appears to be less related to urban zone than the likelihood of participating in international travel. Only the coefficient of intensive transit zone is statistically significant, suggesting that the residents of the zone participate less in national travel than the residents of the central pedestrian zone. Income and education level have a positive association with national travel, which is to be expected. Surprisingly, men seem to participate less in national long-distance travel than women. This may be because of limiting the scope of the study to leisure and family long-distance trips and young adults. In general, the differences in travel behaviour between men and women are small in the study. 3.3.2. Amount of emissions (linear regression) The impact of the studied urban zones on the amount of the emissions among those who participate in long-distance travel is not as clear as the impact on the participation. In the case of international travel (model 1b), only the coefficient of basic transit zone is statistically significant. The negative coefficient suggests that the residents of the pedestrian zone do not only travel more often but also further than the residents of basic transit zone. In addition, the emissions of those who make international trips seem to increase with increasing income and education level. Families with children have lower emissions than singles. In the case of national travel, the studied urban zones have no statistically significant impact on the amount of the emissions among those who participate in national travel. Again, increasing income and education level have increasing impact on the emissions of those who travel. Couples and families with children have lower emissions than singles. 3.3.3. Other tested variables In addition to the regression models presented in Table 4, we tested models with several other variables: car-ownership, type of residence, private yard, weekly working hours and participation in business trips (dummy variable). The tested models are included in the SI. Interestingly, car-ownership, type of residence and private yard were in most cases statistically insignificant (SI Models 3 to 5). We tested their impact with separate models. However, consistent with the descriptive section above, people who work 35 to 40 h are more likely to travel internationally than people who work less and people who work more (SI Model 1c). It suggests that very high workload may lead to reduction of participation in leisure and family travel. Results were similar for emissions (SI Model 1d) and in the case of national travel (SI Models 2c and 2d). The business trip variables (national and international business trips separately) were not statistically significant in any cases.

3.3. Regression analyses Regression analyses support the finding that the emissions caused by international travel are highest for the residents of the central pedestrian zone. When income and other socioeconomic variables are controlled, urban zone has an independent impact on the emissions. The residents of public transit zones and sub-centers cause lower emissions than the pedestrian zone residents. However, there is no statistically significant difference between car zone and the central pedestrian zone. In the regression models the fringe of the central pedestrian zone has been combined with the central pedestrian zone due to their structural similarity, and to achieve larger group size and higher statistical significance (see the method section). Table 4 presents the main results from the regression analyses. We ran two regression models: (1) logistic model for participation in emissions (traveled = 1, did not travel = 0) and (2) an ordinary least squares (OLS) model for emissions (including only respondents with emissions). Table 4 provides model estimations for international and national travel separately. Residuals of the models were not spatially autocorrelated when tested with Moran's I statistic, which suggests that explanatory variables included in the models adequately explain spatial variation of the outcome variables. The regression coefficients are not standardized in the table.

4. Discussion The study was set to investigate the travel patterns and factors of differences in these patterns among young adults living in different urban zones of HMA in Finland. Two objectives were set: 1) to compare the GHG emissions related to local, national and international leisure travel between residents of different urban zones of HMA, and 2) to identify the key factors of the differences in travel-related GHG emissions, specifically whether residential location remains a significant factor when socio-demographic and life situation variables are controlled. The data were collected with a softGIS survey and cover all modes of travel and local, national and international trips. The data collection format allowed accounting for location of residences within the urban region. Such level of detail has often been unavailable in previous similar studies (e.g. Holz-Rau et al., 2014; Reichert et al., 2016) and thus the study makes an important contribution. Another key value lies in

3.3.1. Participation in national and international travel (logistic regression) Higher income and living in the central pedestrian zones are related to higher likelihood of travelling internationally in the past year. However, the difference between the pedestrian zone and car zone is not statistically significant. The effect of urban zone is similar in size as the impact of income level. In addition, education seems to increase the likelihood of international travel to some extent. Interestingly, 137

Journal of Transport Geography 68 (2018) 130–141

M. Czepkiewicz et al.

Table 4 Results of two-part regression models: logistic model for participation in emissions from international travel (1a), OLS regression model for emissions (ln) from international travel (1b), logistic model for participation in emissions from national travel (2a), OLS regression model for emissions (ln) from national travel (2b). Models 1b and 2b include only those participating in international travel and national travel, respectively. Coefficients statistically significant at 95% level (p < .05) in bold font. International travel

Income

Education

Household

Gender Urban zone

Constant

National travel

Model 1a

Model 1b

Model 2a

Model 2b

Logistic

ln_emissions

Logistic

ln_emissions

Log likelihood R2 or pseudo R2

−354.1 0.085

0.102

−221.5 0.102

0.054

< 1500 EUR 1500–3000 EUR 3000–4500 EUR 4500–6500 EUR > 6500 EUR Basic and secondary Post-secondary Graduate and post-grad. Single Couple living together Families with children Other Female Male Central pedestrian zones Intensive public transportation zone Pedestrian zones of the sub-centers Basic public transportation zone Car zone

Coeff. −0.52 0.47 – 1.01 1.17 – 0.58 0.28 – −0.09 0.26 −0.19 – −0.26 – −0.57 −1.03 −0.61 −0.20 1.40

P > |z| 0.095 0.078 – 0.001 0.000 – 0.024 0.277 – 0.712 0.332 0.740 – 0.184 – 0.037 0.001 0.027 0.571 0.003

Coeff. −0.09 −0.05 – 0.16 0.33 – 0.41 0.60 – 0.11 −0.42 0.67 – −0.11 – −0.23 −0.15 −0.28 −0.09 7.37

P > |z| 0.658 0.761 – 0.270 0.028 – 0.004 0.000 – 0.440 0.004 0.042 – 0.250 – 0.102 0.395 0.044 0.551 0.000

Coeff. −1.37 −0.49 – 0.12 0.28 – 0.68 0.47 – 0.43 −0.02 1.13 – −0.75 – −0.76 −0.55 −0.36 −0.52 3.48

P > |z| 0.001 0.203 – 0.774 0.550 – 0.039 0.149 – 0.236 0.961 0.301 – 0.005 – 0.038 0.207 0.360 0.245 0.000

Coeff. −0.36 −0.19 – 0.29 0.45 – 0.26 0.20 – −0.28 −0.33 −0.32 – 0.03 – −0.08 −0.06 0.12 0.02 6.33

P > |z| 0.054 0.154 – 0.028 0.001 – 0.050 0.131 – 0.028 0.012 0.267 – 0.733 – 0.497 0.678 0.298 0.865 0.000

When analyzing the impacts of various factors with regression analysis, we found that the urban zone has an effect on international long-distance travel independently from socioeconomic characteristics. Particularly, the residents of the central pedestrian zone participate more in international travel compared to the residents of public transport zones and sub-centers, controlling for their income, education level and household type. The effect of urban zones on the amount of emissions among those participating was weaker. Both income and education level were positively related to participation in and amount of national and international travel-related emissions. This hints at a cultural rather than solely economic character of the relationship. Household type does not seem to affect participation in long-distance travel, but families with children are likely to cause less emissions than singles among those who participate. The results are in line with previous studies that have used regression analysis to examine the relationship between urban structure and long-distance travel within metropolitan regions (Holden and Norland, 2005; Næss, 2006, 2016; Holden and Linnerud, 2011). For example, Næss (2016) found that the number of flights decreases with increasing distance to the main city center, and Holden and Linnerud (2011) found that the long-distance leisure travel by plane (kWh/person/year) increases with residential density. In addition to the urban zones, the impact of car-ownership, apartment and private yard were tested separately, but found not to influence long-distance travel in the study. However, the study supported the authors' hypothesis related to working hours: the amount of working first increases long-distance travel, but then decreases it – people who work most travel less than people who work moderately. One possible explanation is that leisure travel is constrained not only by money but also by time and people who work a lot, have less time to travel. The other explanations might relate to differences in lifestyle orientations (e.g. Ohnmacht et al., 2009; van Acker et al., 2010). The study does not look into more specific hypotheses related to causal mechanism behind the increased amount of long-distance travel by residents of certain urban zones. To guide future research on the topic,

the longer time frame for national and international travel than in most of existing travel datasets (Reichert et al., 2016; Ottelin et al., 2014), leading to more reliable insight into yearly amounts of long-distance travel, and the less problematic presence of multiple zeroes in the datasets. A similar 12-month time span has been used before as well (Høyer and Holden, 2003; Næss, 2006, 2016; Ornetzeder et al., 2008; Brand and Boardman, 2008; Brand and Preston, 2010; Bruderer Enzler, 2017), but has not been very common. Furthermore, the survey includes information for both local and longer distance travel for the same respondents, thus allowing travel pattern analyses over all trip types. Another important strength of the study is the use of reported locations of trip destinations and calculation of distances based on road and railroad networks for relevant travel modes, which increases data accuracy. Finally, the inclusion of the infrastructure part to the GHG assessment is a value-adding factor, since the indirect emissions have been found important (Chester and Horvath, 2009; Chester et al., 2013), but are still rarely included in any transport sector assessments. In the study, we found that emissions from local travel are the lowest on average among young residents of centrally located and densely built urban zones - a result that is congruent with previous studies conducted in Nordic countries (Næss, 2012) and internationally (Ewing and Cervero, 2010). The study also found that the same group of residents generates high overall GHG emissions due to elevated international travel activity. This is also in line with results of previous research (Ottelin et al., 2014; Reichert et al., 2016). However, in this study the key finding is that it is the international long-distance trips which dominate the travel-related GHG emissions, which holds true in all travel zones among the studied sample. Furthermore, even national trips cause more GHG emissions than local trips across most urban zones. Using GWP20 (or other shorter time horizon) metric instead of GWP100 would also have significantly further increased the role of air travel due to SLCFs like contrail and cirrus showing much more strongly with metrics with shorter time-horizons, as shown by e.g. Aamaas et al. (2013) and Peters et al. (2011). For other transport modes the metric time-horizon is not similarly important. 138

Journal of Transport Geography 68 (2018) 130–141

M. Czepkiewicz et al.

emissions by these travel modes. Furthermore, the choice of a 1.2 multiplier to account for detours due to interchanges was a rough estimate. More basic research on how actual routes and emissions divert from shortest routes would be necessary to inform future research. Potential biases bring uncertainty to the direct comparisons between GHG emissions from different travel distances and travel modes. However, again they should be similar enough for such general comparisons, as carried out in this study. Another type of limitation is imposed by the GHG assessment scope and impact factors adopted in the study. First, the utilized GWP100 metric has been criticized e.g. for the treatment of time and for the omission of the SLCFs, which form an important share of the aviation climate impact (e.g. Aamaas et al., 2013; Peters et al., 2011; Lee et al., 2010). We corrected for the latter problem by using SLCFs corrected GWP100 factors for aviation. This leads to air travel vastly dominating the overall emissions, but still the GWP100 metric emphasizes the role of the SLCFs less than several suggested alternative metrics, particularly those using shorter time-perspectives (Aamaas et al., 2013; Lee et al., 2009 & Lee et al., 2010; Borken-Kleefeld et al., 2013). Thus, the share of air travel in the overall estimates is still likely underestimated rather than overestimated. There is also much higher uncertainty associated with the SLCFs than the LLGHGs (Aamaas and Peters, 2017; Peters et al., 2011; Lee et al., 2010). Second, while a valuable addition, the inclusion of the infrastructure part of the GHG factors brings additional uncertainty. There are uncertainties related to the source study of Chester and Horvath (2009) itself and assuming the results to be applicable in Finnish conditions and over international trips made by the participants increases the uncertainty level. To reduce this uncertainty, we ran the same assessments using only the more certain direct emissions and fuel production emissions. The results largely comply with the results obtained with the infrastructure included. Finally, emissions per PKT are highly dependent on the occupancy rate of the vehicle. Since we do not have data for the occupancy rate, we utilized averages from the sources explained in Section 2.4. The uncertainty relates the most strongly to car travel, resulting in a potential exaggeration of the differences between the residents of different zones due to differences in household sizes, particularly between the pedestrian zone, where adult households dominated, and car-oriented zones with families with children as the main household type. The assumed average occupancies might be too high for adult households and too low for families with children. As with much of studies that use PPGIS for data collection, spatial accuracy of markings brings about uncertainty (Brown, 2012; Brown and Kyttä, 2014). Accuracy of markings depends on mapping tool characteristics, such as map zoom level, question wording and tool usability, as well as participant characteristics, such as engagement in the survey, browsing skills, map literacy and familiarity with the study area (Brown, 2012). Spatial inaccuracy of markings may result in errors in matching residential locations with urban zones and errors in distance calculation. Since lack of accuracy may result in both overestimation and underestimation of distance, the errors in both directions are likely to cancel out each other, and there is no reason to assume the errors to be higher for certain samples than for the others. The only known case of likely overestimation of trip distances are of ferry trips to Estonia and Sweden. The destinations were often marked much further inland than the port cities of Tallinn and Stockholm, most probably because of the small-scale map that was used for marking international trips (World map at Google Maps zoom level 2). Because the study covered only one urban region, the results may not be generalizable to other cities and regions. The Helsinki metropolitan area has specific characteristics relevant for the study. When compared globally, the region has high levels of livability and environmental quality, a high purchasing power of the population, and a low level of socioeconomic segregation. The region also has a monocentric character with the vital role of the urban core as the cultural and economic center and an attractive place to live in. This is different from

we provide a brief overview of potential explanations. Certain lifestyles and preferences may be best realized in cities and especially dense and diverse neighborhoods and that the same lifestyles may be linked to a need or wish for long-distance travel (both business and private) (HolzRau et al., 2014). Decisions to have a yard or to buy a car (and to choose a place to live) may be determined by the same values and preferences that influence leisure travel choices (Holden and Linnerud, 2011) and certain “urban” and “cosmopolitan” lifestyles, e.g. the ones that are popular among students and academics, are linked in similar ways to willingness to fly and to live in inner parts of cities (Næss, 2006). On the other hand, dense neighborhoods in big metropolises may facilitate and support creation of cosmopolitan cultures and lifestyles that are related to long-distance travel, as tentatively suggested by Holden and Linnerud (2011). The relationship between urban form and long-distance travel may also be economic in character. According to the “rebound effect hypothesis”, people who travel less on an everyday basis, or don't own and pay for cars, may use the remaining parts of their time travel budgets or financial budgets on leisure travel (Holden and Linnerud, 2011; Strandell and Hall, 2015; Heinonen et al., 2013b; Ottelin et al., 2017). Another possible causal link between urban structure and travel is related to the compensation hypothesis. Poor access to green spaces and leisure opportunities in a residential environment may lead to more long-distance travel, for instance to second homes or for holidays (Strandell and Hall, 2015). Compensation behavior may be linked not only to green areas but also to poor environmental quality and such effects as noise, crowding, hectic atmosphere, and stress (see Næss, 2006; Maat and de Vries, 2006). Strandell and Hall (2015) suggest that, while access to green areas seems to be unrelated, compensation may occur when perceived quality is poor due to noise, congestion, and rushing in a dense environment. 4.1. Uncertainties The study has uncertainties and limitations related to the data collection, GHG assessment method and restricted scope. Differences in the time frame for local trips and long-distance trips may influence comparisons of absolute values (reported in Section 3.1). For local trips, the instruction for the respondent did not specify a time frame (“Please mark locations that you have been frequently visiting within Helsinki metropolitan area.”). For longer distance trips, the respondents were advised to mark all trips within the last 12 months, which likely is a long-enough time frame to catch the relevant travel patterns. While being an improvement compared to studies measuring long-distance travel over shorter periods (e.g. Reichert et al., 2016) or up to a certain number of trips per year (e.g. Bruderer Enzler, 2017), the measurement is not free from biases due to recall errors, potentially both over- and under-estimating trip numbers. Local trips were also mapped using a closed list of most common trip categories. The data collection method may thus have led to underestimation of the number and diversity of local trips, as less frequent and less typical trips could possibly form an important share of all trips. Moreover, emission estimation relies on the number of locations marked on the map, which may be influenced by the level of respondents' engagement in mapping activity and level of their mapping skills (Brown, 2012). However, the bias pertains to individual participants, and there is no known indication that it systematically relates to the variables of interest of this study. To provide some indication of accuracy, wee validated softGIS data for car travel using annual mileages and fuel efficiencies of cars available in households, and even if some differences appear, the validation suggests there is no consistent downward bias in the softGIS measurement of local trips by car. There could still exist such a bias for other local travel modes, but they are less significant for the GHG emission analysis which is the nexus of this study. Another source of bias is related to distance estimation in long-distance trips. In case of trips by plane and ferry, we took a geodetic distance between origins and destinations, assuming that the shortest routes are taken, which may lead to underestimation of 139

Journal of Transport Geography 68 (2018) 130–141

M. Czepkiewicz et al.

of Iceland Research Fund and the Marketta Kyttä's tenure funding.

other cities, especially in the United States, where urban cores are still often of lower environmental quality, attractiveness and socioeconomic status than suburbs (e.g. Ehrenhalt, 2013). Due to these specific characteristics, the results of our study may not be generalizable to cities with different urban structures and socioeconomic characteristics. The results may also be culturally limited to the Nordic context and may not be replicable in cities in other parts of Europe and the world. In addition to the geographical restriction of the HMA there was also age group restriction. The results might be thus best generalizable to young adults. However, similar patterns were found in a larger scale study by Ottelin et al. (2014) where wider age groups and wider geographical areas were studied. This suggests that the results might apply to the wider population as well.

Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.jtrangeo.2018.02.008. References Aamaas, B., Peters, G., 2017. The climate impact of Norwegians' travel behavior. Travel Behaviour and Society 6, 10–18. Aamaas, B., Borken-Kleefeld, J., Peters, G., 2013. The climate impact of travel behavior: a German case study with illustrative mitigation options. Environ. Sci. Pol. 33, 273–282. van Acker, V., van Wee, B., Witlox, F., 2010. When transport geography meets social psychology: toward a conceptual model of travel behaviour. Transp. Rev. 30, 219–240. http://dx.doi.org/10.1080/01441640902943453. Ala-Mantila, S., Ottelin, J., Heinonen, J., Junnila, S., 2016. To each their own? The greenhouse gas impacts of intra-household sharing in different urban zones. J. Clean. Prod. 135, 356–367. Alcock, I., White, M.P., Taylor, T., Coldwell, D.F., Gribble, M.O., Evans, K.L., ... Fleming, L.E., 2017. “Green” on the ground but not in the air: pro-environmental attitudes are related to household behaviours but not discretionary air travel. Glob. Environ. Chang. 42, 136–147. http://dx.doi.org/10.1016/j.gloenvcha.2016.11.005. Borken-Kleefeld, J., Fuglestvedt, J., Berntsen, T., 2013. Mode, load, and specific climate impact from passenger trips. Environ. Sci. Technol. 47, 7608−7614. Boucher, J.L., 2016. Culture, carbon, and climate change: a class analysis of climate change belief, lifestyle lock-in, and personal carbon footprint. Socijalna Ekologija 25 (1), 53–80. http://dx.doi.org/10.17234/SocEkol.25.1.3. Bramley, G., Power, S., 2009. Urban form and social sustainability: the role of density and housing type. Environ. Plan. B 36 (1), 30–48. http://dx.doi.org/10.1068/b33129. Brand, C., Boardman, B., 2008. Taming of the few-the unequal distribution of greenhouse gas emissions from personal travel in the UK. Energy Policy 36 (1), 224–238. Brand, C., Preston, J.M., 2010. ‘60-20 emission’—the unequal distribution of greenhouse gas emissions from personal, non-business travel in the UK. Transp. Policy 17 (1), 9–19. Brown, G., 2012. An empirical evaluation of the spatial accuracy of public participation GIS (PPGIS) data. Appl. Geogr. 34 (2), 289–294. http://dx.doi.org/10.1016/j.apgeog. 2011.12.004. Brown, G., Kyttä, M., 2014. Key issues and research priorities for public participation GIS (PPGIS): a synthesis based on empirical research. Appl. Geogr. 46, 122–136. Brownstone, D., Golob, T.F., 2009. The impact of residential density on vehicle usage and energy consumption. J. Urban Econ. 65 (1), 91–98. Bruderer Enzler, H., 2017. Air travel for private purposes; an analysis of airport access, income and environmental concern in Switzerland. J. Transp. Geogr. 61 (January), 1–8. http://dx.doi.org/10.1016/j.jtrangeo.2017.03.014. Center for International Earth Science Information Network, 2013. Global Roads Open Access Data Set, Version 1 (gROADSv1). Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC). Columbia University, and Information Technology Outreach Services, ITOS - University of Georgia. doi:https://doi.org/10.7927/ H4VD6WCT. Accessed 10th September 2016. Chapman, L., 2007. Transport and climate change: a review. J. Transp. Geogr. 15 (5), 354–367. Chester, M., Horvath, A., 2009. Environmental assessment of passenger transportation should include infrastructure and supply chains. Environ. Res. Lett. 4, 024008 (8pp.). Chester, M., Pincetl, S., Elizabeth, Z., Eisenstein, W., Matute, J., 2013. Infrastructure and automobile shifts: positioning transit to reduce life-cycle environmental impacts for urban sustainability goals. Environ. Res. Lett. 8 (1), 015041. Chitnis, M., Sorrell, S., Druckman, A., Firth, S.K., Jackson, T., 2013. Turning lights into flights: estimating direct and indirect rebound effects for UK households. Energy Policy 55, 234–250. Ehrenhalt, A., 2013. The Great Inversion and the Future of the American City. (Vintage). Ewing, R., Cervero, R., 2010. Travel and the built environment: a meta-analysis. J. Am. Plan. Assoc. 76 (3), 265–294. Finnish Environment Institute SYKE, 2010. Travel-related Urban Zones. In: Downloaded from the Open Data Web Service 18.05.2016 using the license Creative Commons By 4.0 International. Fletcher, D., MacKenzie, D., Villouta, E., 2005. Modelling skewed data with many zeros: a simple approach combining ordinary and logistic regression. Environ. Ecol. Stat. 12 (1), 45–54. http://dx.doi.org/10.1007/s10651-005-6817-1. Frändberg, L., Vilhelmson, B., 2011. More or less travel: personal mobility trends in the Swedish population focusing gender and cohort. J. Transp. Geogr. 19 (6), 1235–1244. Haybatollahi, M., Czepkiewicz, M., Laatikainen, T., Kyttä, M., 2015. Neighbourhood preferences, active travel behaviour, and built environment: an exploratory study. Transport. Res. F: Traffic Psychol. Behav. 29. http://dx.doi.org/10.1016/j.trf.2015. 01.001. Heinonen, J., Jalas, M., Juntunen, J.K., Ala-Mantila, S., Junnila, S., 2013a. Situated lifestyles: I. How lifestyles change along with the level of urbanization and what the greenhouse gas implications are—a study of Finland. Environ. Res. Lett. 8 (2). Heinonen, J., Jalas, M., Juntunen, J.K., Ala-Mantila, S., Junnila, S., 2013b. Situated lifestyles: II. The impacts of urban density, housing type and motorization on the greenhouse gas emissions of the middle income consumers in Finland. Environ. Res. Lett. 8 (3).

4.2. Future research Future studies should seek to replicate the results in urban regions with characteristics both similar to and different from the Helsinki metropolitan area, and in distinct cultural and economic contexts. Researchers should also focus on identifying causal links between urban form and travel-related emissions and explaining rationales behind travel behavior. For this purpose, using more sophisticated modeling methods, such as structural equation modeling (SEM) and mixedmethods research designs, could be instrumental. Finding causality between urban form and long-distance travel is especially important from a policy perspective, as doing so may challenge urban planning policies to some extent, such as a compact city (Holden and Norland, 2005; Holden and Linnerud, 2011). In this context, research focused on compensation hypotheses is especially relevant. Finding which characteristics of urban environment increase long distance travel would probably not undermine densification policies, but rather help to refine them. For instance, such results could instruct urban planners on which local environment qualities should be in place to make people spend more leisure time in the same local area rather than leaving the city for long-distance trips. A possible venue for such future research is looking in more detail into urban structural characteristics, which may have both positive and negative influence on urban livability and residents' evaluations of environmental quality (e.g. Bramley and Power, 2009; Kyttä et al., 2015; Mouratidis, 2017), and connecting these with prevalence of escape trips. On the other hand, research could also concentrate on trip purposes and perceived well-being benefits associated with long-distance travel. So far, we know too little about the nature of the long-distance trips to make conclusions about the role of the urban environment in motivating them. Studies should also look more deeply into economic, cultural, and psychological factors behind long-distance travel. Future studies would benefit from including ‘soft’ variables such as personal attitudes, values, and lifestyles in the equation (van Acker et al., 2010). Specifically, the role of cosmopolitan lifestyles (Reichert et al., 2016) and class cultures (Boucher, 2016) should be studied in more detail. Interrelationships between residential location, travel patterns, and pro-environmental attitudes is another potential line of investigation. A discrepancy between pro-environmental attitudes and air travel has been reported in several studies (Holden and Linnerud, 2011; Alcock et al., 2017). It is possible that such an attitude-behavior gap is contextually associated with an urban environment. Rejection of causal links between urban form and long-distance travel would also have important policy implications. Establishing that long-distance travel behavior has only cultural and psychological, rather than environmental determinants, would emphasize the role of education, social marketing and other strategies like carbon taxation or personal quotas. Acknowledgments The research presented in the article was funded by the University 140

Journal of Transport Geography 68 (2018) 130–141

M. Czepkiewicz et al.

Holden, E., Linnerud, K., 2011. Troublesome leisure travel: the contradictions of three sustainable transport policies. Urban Stud. 48, 3087–3106. Holden, E., Norland, I.T., 2005. Three challenges for the compact city as a sustainable urban form: household consumption of energy and transport in eight residential areas in the greater Oslo region. Urban Stud. 42 (12), 2145–2166. Holz-Rau, C., Scheiner, J., Sicks, K., 2014. Travel distances in daily travel and long-distance travel: what role is played by urban form? Environ Plan A 46 (2), 488–507. Høyer, K.G., Holden, E., 2003. Household consumption and ecological footprints in Norway – does urban form matter? J. Consum. Policy 26 (3), 327–349. Kahila, M., Kyttä, M., 2009. SoftGIS as a Bridge-Builder in Collaborative Urban Planning in: Planning Support Systems Best Practice and New Methods. Springer, pp. 389–411. Kwan, M.-P., 2012. How GIS can help address the uncertain geographic context problem in social science research. Ann. GIS 18 (4), 245–255. http://dx.doi.org/10.1080/ 19475683.2012.727867. Kyttä, M. 2011, SoftGIS methods in planning evaluation, in Evaluation for Sustainability and Participation in Planning, pp. 334e354. Kyttä, M., Broberg, A., Haybatollahi, M., Schmidt-Thomé, K., 2015. Urban happiness: context-sensitive study of the social sustainability of urban settings. Environ. Plan. B 47, 1–24. http://dx.doi.org/10.1177/0265813515600121. Lee, D., Fahey, D., Forster, P., Newton, P., Wit, R., Lim, L., Owen, B., Sausen, R., 2009. Aviation and global climate change in the 21st century. Atmos. Environ. 43, 3520–3537. Lee, D.S., Pitari, G., Grewe, V., Gierens, K., Penner, J.E., Petzold, A., Prather, M.J., Schumann, U., Bais, A., Berntsen, T., Iachetti, D., Lim, L.L., Sausen, R., 2010. Transport impacts on atmosphere and climate: aviation. Atmos. Environ. 44, 4678–4734. Lenzen, M., Dey, C., Foran, B., 2004. Energy requirements of Sydney households. Ecol. Econ. 49 (3), 375–399. Maat, K., de Vries, P., 2006. The influence of the residential environment on greenspace travel: testing the compensation hypothesis. Environ Plan A 38, 2111–2127. http:// dx.doi.org/10.1068/a37448. Min, Y., Agresti, A., 2002. Modeling nonnegative data with clumping at zero: a survey model for semicontinuous data. J. Iran. Stat. Soc. 1 (May), 7–33. Mindali, O., Raveh, A., Salomon, I., 2004. Urban density and energy consumption: a new look at old statistics. Transp. Res. A Policy Pract. 38 (2), 143–162. Mouratidis, K., 2017. Is compact city livable? The impact of compact versus sprawled neighbourhoods on neighbourhood satisfaction. Urban Stud. 4209801772910. http://dx.doi.org/10.1177/0042098017729109. Næss, P., 2006. Urban Structure Matters: Residential Location, Car Dependence and Travel Behaviour. Routledge, Abingdon. Næss, P., 2012. Urban form and travel behavior: experience from a Nordic context. J. Transp. Land Use 5, 21–45. http://dx.doi.org/10.5198/jtlu.v5i2.314. Næss, P., 2016. Urban planning: Residential location and compensatory behaviour in three Scandinavian cities. book section in In: Rethinking Climate and Energy Policies. Newman, P., Kenworthy, J., 1989. Cities and Automobile Dependence: An International Sourcebook. Gover, Aldershot. Newman, P., Kenworthy, J., 1999. Sustainability and Cities: Overcoming Automobile

Dependence. Island Press. Norman, J., MacLean, H., Kennedy, C., 2006. Comparing high and low residential density: life-cycle analysis of energy use and greenhouse gas emissions. J. Urban Plan. Dev. 132 (1), 10–21. Ohnmacht, T., Götz, K., Schad, H., 2009. Leisure mobility styles in Swiss conurbations: construction and empirical analysis. Transportation 36, 243–265. http://dx.doi.org/ 10.1007/s11116-009-9198-8. Ornetzeder, M., Hertwich, E.G., Hubacek, K., Korytarova, K., Haas, W., 2008. The environmental effect of car-free housing: a case in Vienna. Ecol. Econ. 65 (3), 516–530. OSF (Official Statistics of Finland), 2016. Population structure. ISSN=1797-5395. In: Helsinki: Statistics Finland [referred: 18.5.2016], Access method: http://www.stat. fi/til/vaerak/index_en.html. Ottelin, J., Heinonen, J., Junnila, S., 2014. Greenhouse gas emissions from flying can offset the gain from reduced driving in dense urban areas. J. Transp. Geogr. 41, 1–9. Ottelin, J., Heinonen, J., Junnila, S., 2017. Rebound effect for reduced car ownership and driving. In: Kristjánsdóttir, S. (Ed.), Nordic Experiences of Sustainable Planning: Policy and Practice. Routledge. Peters, G.P., Aamaas, B., Lund, M.T., Solli, C., Fuglestvedt, J.S., 2011. Alternative “Global Warming” metrics in life cycle assessment: a case study with existing transportation data. Environ. Sci. Technol., 45 (20), 8633–8641. Reichert, A., Holz-Rau, C., 2015. Mode use in long-distance travel. J. Transp. Land Use 8 (2), 87–105. http://dx.doi.org/10.5198/jtlu.2015.576. Reichert, A., Holz-Rau, C., Scheiner, J., 2016. GHG emissions in daily travel and longdistance travel in Germany–social and spatial correlates. Transp. Res. Part D: Transp. Environ. 49, 25–43. Ristimäki, M., Kalenoja, H., Tiitu, M., 2011. Yhdyskuntarakenteen vyöhykkeet. Vyöhykkeiden kriteerit, alueprofiilit ja liikkumistottumukset. Liikenne-ja viestintäministeriön julkaisuja 15 (2011), 2011–2015. Ristimäki, M., Tiitu, M., Kalenoja, H., Helminen, V., Söderström, P., 2013. Yhdyskuntarakenteen vyöhykkeet Suomessa-Jalankulku. joukkoliikenne-ja autovyöhykkeiden kehitys vuosina 1985–2010. Salonen, M., Broberg, A., Kyttä, M., Toivonen, T., 2014. Do suburban residents prefer the fastest or low-carbon travel modes? Combining public participation GIS and multimodal travel time analysis for daily mobility research. Appl. Geogr. 53, 438–448. http://dx.doi.org/10.1016/j.apgeog.2014.06.028. Söderström, P., Schulman, H., Ristimäki, M., 2015. Urban form in the Helsinki and Stockholm city regions: development of pedestrian, public transport and car zones. In: Reports of the Finnish Environment Institute. 2015. Strandell, A., Hall, C.M., 2015. Impact of the residential environment on second home use in Finland – testing the compensation hypothesis. Landsc. Urban Plan. 133, 12–23. U.S. Environmental Protection Agency, 2008. Direct Emissions from Mobile Combustion Sources. US Environmental Protection Agency Office of Air and Radiation May 2008. VTT Technical Research Centre of Finland, 2012. LIPASTO – a calculation system for traffic exhaust emissions and energy consumption in Finland. VTT Technical Research Centre of Finland, 2016. LIPASTO – a calculation system for traffic exhaust emissions and energy consumption in Finland. http://lipasto.vtt.fi (accessed 12.9.2016).

141