Urbani izziv, volume 31, no. 1, 2020
112
UDC: 656.025.2:796.61(497.451.1)
DOI: 10.5379/urbani-izziv-en-2020-31-01-005
Received: 17 Apr. 2020
Accepted: 13 July 2020
Simon KOBLAR
Luka MLADENOVIČ
Calculating the speed of city bus trips:  
The case of Ljubljana, Slovenia
In promoting the use of public transport, an understand-
ing of the passengers’ perspective on the provided ser-
vice plays an important role. A series of factors inuence 
people’s selection of transport mode, among which the 
competitiveness of travel time, or travel speed, is vital. 
anks to the widespread use of electronic payment sys-
tems, data collected through user validation can be used 
to calculate this speed. us, the actual trips made can 
be used to estimate their speed. is study focused on 
the Ljubljana bus system to analyse all trips made on a 
typical day. e input and output trip data were used to 
calculate the distance travelled, and the time and speed 
of the trips. In addition, an estimate was also made of 
how quickly the distances travelled by bus could have 
been travelled by bicycle or on foot. e ndings showed 
that the speed of the bus trips analysed depends on the 
length of the journey: it increases with longer journeys. 
Bicycles are generally faster for all distances, but they be-
come a less acceptable choice for longer distances. With 
regard to distances shorter than 2 km, in terms of speed, 
walking is competitive on only a few routes. e analyses 
performed using the data collected through the electronic 
service payment system provided useful insight into the 
eciency of the public transport system from the passen-
ger perspective, which in the future may prove useful in 
planning system improvements.
 
Keywords: public transport, travel speed, eective travel 
speed, electronic payment system, speed comparison
Urbani izziv, volume 31, no. 1, 2020
113
1 Introduction
Understanding the residents’ travel habits and reasons for them 
is an important factor in promoting sustainable mobility. e 
goals of sustainable mobility measures are oen directed to-
wards changing people’s travel habits, especially reducing the 
use of cars and promoting the use of public transport, cycling, 
and walking as dierent modes of making daily trips. People’s 
decisions to use public transport are heavily inuenced by its 
quality  (Vanhanen  & Kurri,  2007). Studies of travel habits 
examine the factors inuencing the choice of travel mode or 
the indicators dening how a public transport system operates. 
e quality indicators of a public transport system can be di-
vided into two major categories: the transit capacity and the 
quality of the actual service provided  (KFH Group,  2013). 
Quality of service is dened using user perceptions or actual 
numerical measurements (Carreira et al., 2014). If the service is 
of good quality, then frequency, availability, travel time, price, 
and sta quality are especially important in deciding to use 
public transport  (Stradling et  al.,  2007). e key indicators, 
which are also important factors in selecting the travel mode, 
are the speed and consequently the time the user spends to 
make a trip. ere are only a few Slovenian studies in this 
area and the ones that do exist do not provide detailed insight 
into the conditions that inuence the passengers’ motivation 
to use public transport (Statistični urad RS, 2017; Ljubljanski 
potniški promet, 2019). Travel time is one of the most impor-
tant elements of public transport quality (KFH Group, 2013) 
because all the other factors inuencing the choice of travel 
mode only come to the fore when the user is provided with a 
competitive selection of various travel modes in terms of travel 
times. Longer travel times  (e.g.,  of commuting to work) are 
directly connected with reduced user satisfaction  (Loong  & 
El-Geneidy,  2016), as well as poorer wellbeing and social in-
clusion (Morris & Guerra, 2015).
Various methods are used to calculate public transport speed. 
e speed over a specic stretch, including all stops and delays, 
is referred to as commercial speed. is indicator is primarily 
important from the operator’s point of view because it makes 
it possible to calculate a vehicle’s travel time on a line, set up 
timetables and drivers’ schedules, and eectively distribute ve-
hicles across the system. From the passengers’ point of view, 
commercial speed is not enough because they compare the 
travel times of various transport modes from a door-to-door 
perspective. More important for them are the time and speed 
that also include the time of reaching the station, waiting, 
in-vehicle travel, any transfers, and ultimately reaching the des-
tination (Munizaga et al., 2016; Constantinescu et al., 2018). 
is speed is referred to as the eective total travel speed below.
Data collected through passenger validation enabled by digi-
talized payment systems provide great potential for acquiring 
data on and analysing these speeds. Such data allow a much 
better understanding of passenger travel habits and it also 
makes sense to use them in improving the public transport 
systems (Schmöcker, 2016). Smart card data can also be used 
to calculate the key indicators of a system’s operation (Trépani-
er & Morency, 2016), as well as conduct many other analyses 
in addition to those focusing on travel speeds (Jang, 2010).
is article presents a method for analysing the speed of public 
transport trips in Ljubljana using the data on the trips actually 
made. e study examines the time parameters of trips without 
the analyses of perceived times. It proceeds from the hypothesis 
that the available payment system and timetable data can be 
used to determine the speed of actual public transport trips 
that is more accurate than the data available to date. e sec-
ond part of the study compares the public transport travel 
speeds with the speeds of traveling the same routes by bicycle 
or on foot. Comparisons of individual travel modes in a city 
are a frequent research topic (Ellison & Greaves, 2011; Ander-
sen, 2014), but most of these are unsystematic. e literature 
review revealed no study that would provide a comparison 
between a public transport mode and cycling based on a suf-
ciently large sample and comparable routes. Based on the 
available data on the relatively short distance of an average 
trip, this study proceeds from the hypothesis that an average 
public transport trip would take less time by bicycle.
1.1 The case of Ljubljana
Public transport in Ljubljana is operated by Ljubljanski Pot-
niški Promet (hereinaer: LPP), which carries nearly 40 mil-
lion passengers a year. In recent years, the number of passengers 
has been falling despite many improvements to the service and 
passenger comfort, such as revamping the bus eet and the 
bus arrivals system, improving the quality of bus stops, and 
introducing separate bus lanes on some arterial roads. e 
main reason for the falling number of passengers is not entire-
ly clear (Ljubljanski potniški promet, 2019). e accessibility 
of public transport is good within the city perimeter  (Ga-
brovec  & Bole,  2006; Kozina,  2010; Gabrovec  & Razpotnik 
Visković, 2012, 2018; Tiran et al., 2015).
Travel times have been poorly studied to date. Celcer (2009) 
analysed the travel times on selected lines and compared them 
to cars, but she did not calculate the travel speeds. She es-
tablished that travel times for cars were signicantly shorter 
on all the routes studied. Travel times on specic lines were 
also studied by Šabič (2015), but he only calculated the com-
mercial speed, which does not take into account waiting and 
Calculating the speed of city bus trips: The case of Ljubljana, Slovenia
Urbani izziv, volume 31, no. 1, 2020
114
walking. Similarly, LPP also only measures the commercial 
speed  (Šmajdek,  2011). Vehicle tracking data were used to 
calculate the travel speed on line  1, which exceeds  22  km/h 
throughout the day  (Čelan  & Lep,  2020). Low travel speed 
as a key problem in public transport has also been highlight-
ed in strategic documents (Milovanović, 2017; Gojčič, 2018), 
in which, however, no current or target values are provided, 
which is most likely the result of this topic being understudied.
e electronic payment system, which Ljubljana introduced 
in  2010, has good potential for analysis. When entering the 
bus, every passenger validates their card or uses the Urbana 
app on their smart phones to pay for the fare. e validation 
data are sent to the central server together with the informa-
tion on the bus stop retrieved from the Automatic Vehicle 
Location (AVL) system (Šmajdek, 2011). Except for keeping 
records of the total number of passengers for the annual re-
ports, these types of data, except for certain exceptions (Kor-
en,  2016; Koblar,  2017; Koblar  & Žebovec,  2018), have not 
been analysed in detail. However, they proved to be very useful 
in analysing user travel patterns (Koblar & Žebovec, 2018; Ko-
blar & Mladenovič, 2020) and planning potential changes to 
the network (Koblar, 2017).
2 Methods
e payment system data were analysed to calculate the travel 
times. Because only the boarding bus stop is recorded in the 
payment system, one of the challenges was determining the 
alighting stops and merging individual trips into a journey. A 
trip refers to a ride on an individual line validated in the pay-
ment system. A journey refers to one or several trips together 
by taking account the boarding stop of the rst trip and the 
alighting stop of the last trip. ese data provided the basis 
for further analyses.
2.1 Determining the alighting stops and 
calculating the travel times
Travel times and speeds were analysed based on the trips made 
and recorded in the payment system used by LPP.  e  2015 
and  2016 validation data retrieved were rst used to select a 
typical day on which an average number of trips (validations) 
were made, the weather was nice (no rain) and there were no 
school holidays, roadblocks or other special events. Wednes-
day, 18 May 2016, was selected, with 142,181 trips recorded.
Because most public transport payment systems are designed 
so that they only record the entry into the vehicle, just like this 
one, a considerable number of authors have so far sought to de-
termine the alighting stops (Cui, 2006; Trépanier et al., 2007; 
Zhao et  al.,  2007; Farzin,  2008; Lu,  2008; Wang,  2010; Li 
et al., 2011; Wang et al., 2011; Alsger et al., 2016; Mosallanejad 
et  al.,  2019; Yan et  al.,  2019; Assemi et  al.,  2020). To dene 
the alighting stops on individuals’ journeys they generally used 
a simple algorithm that compared two daily trips and took 
account of two criteria: the alighting stop on the rst trip is 
the same as the boarding stop on the next trip and the alighting 
stop on the last trip of the day is the same as the boarding stop 
on the rst trip. In addition to determining alighting stops, 
the reconstruction of journeys also requires merging individ-
ual trips into complete journeys. Here, it is vital to accurately 
determine when a person changes lines and continues their 
journey and when they end it. is can be determined based 
on the distance between the alighting stop on the previous line 
and the boarding stop on the next line and the time between 
alighting and the next boarding (Alsger et al., 2016). Due to 
the lack of appropriate data, most researchers did not check 
the accuracy of their results. Alsger et al. (2016) made an im-
portant step toward improving the algorithms and checking 
the quality of results. ey used the smart card data of the 
South-East Queensland public transport network, in which 
passengers also validate their cards when alighting, to check the 
accuracy of results. By modifying established algorithms and 
including data from the public transport schedules, they man-
aged to additionally improve the quality of origin-destination 
estimation algorithms. Later additional improvements were 
introduced, using more complex methods  (machine learn-
ing) to more successfully determine the alighting stops  (Yan 
et al., 2019; Assemi et al., 2020). Due to its simpler implemen-
tation and satisfactory results, we decided to use the algorithm 
proposed by Alsger et al. (2016).
To determine the alighting stops based on this algorithm, the 
smart card validation data must contain the card identier, 
travel time, and the stop and line used. e data obtained in-
clude all the necessary information; in addition, a bus schedule 
database was obtained that was suitably structured for link-
ing with the validation data. Before running the analysis, trips 
without the required data were eliminated from the database. 
Some trips were part of long-distance (inter-city) lines and so 
were not included in the city public transport schedule, and 
for some the wrong line or stop was recorded. Because the 
alighting stop can only be determined for passengers that took 
more than one trip on the same day, data on users with only 
one trip on a selected day were also eliminated from the data-
base (17,614). e basic conditions for inclusion in the analysis 
were met by 113,985 or 80.2% of all the trips made. A matrix 
of distances between the stops is required to determine the 
alighting stops and transfers. For stops less than 800 m apart, 
the distances were modelled based on the road network, which 
resulted in more accurate calculations. For distances between 
other stops, the Euclidean distance was calculated because the 
S. KOBLAR, L. MLADENOVIČ
Urbani izziv, volume 31, no. 1, 2020
115
calculation for the  840  ×  840 matrix of the stops analysed 
would have taken too much time.
e alighting stops were determined using our own soware, 
which followed the algorithm applied  (Alsger et  al.,  2016). 
e soware rst analyses the consecutive trips of the same 
person and orders them into journeys. One journey can be 
composed of several trips with transfers in between. e poten-
tial alighting stops were determined based on the bus schedule, 
from which the potential alighting stop is selected according 
to the line used. From among the stops selected in the previous 
step, the stop closest to the next boarding stop is dened as 
the alighting stop. To determine the alighting time, the travel 
time between both stops as provided in the bus schedule is 
added to the boarding time. If the next boarding stop is less 
than 800 m away and less than 60 min have passed in between, 
the trip is dened as a transfer; otherwise, it is treated as an 
independent journey. In the event of a transfer, the soware 
continues to analyse the user’s card validations until the last trip 
in the journey. If this is the last trip of the day, the stop closest 
to the boarding stop of the rst trip of the day is selected as 
the alighting stop, and the soware then continues by analys-
ing the next user’s trips. e alighting stops were determined 
for 110,069 or 96.5% of validations that met the conditions for 
inclusion in the analysis. e result of the analysis is a consec-
utively numbered list of trips with additional information on 
the alighting stop and the alighting time. Trips that continued 
with a transfer to the next line also contain information on 
the distance to the next boarding stop. ese data were then 
merged into individual trips, for which the travel times were 
calculated.
2.2 Calculating the average waiting time
One of the factors aecting the travel time is also the time of 
waiting for the bus to arrive. Assuming that passengers arrive 
at the stop randomly, the average waiting time depends on the 
frequency of bus trips on all lines that are heading in the select-
ed direction and are available at the time of travel. erefore, 
the dierence in the travel times of the current, previous, and 
next trips were calculated for the specic line used. If this was 
the rst or last trip of the day, only the dierence to the next 
or previous trip was taken into account. e same method was 
used to calculate the waiting times for other lines that could 
have been used between the two selected stops. In this, only 
the lines on which the nearest scheduled departure was less 
than  5  min before or aer the actual trip made were taken 
into account. To calculate the average waiting time, the waiting 
times on individual lines were converted into frequencies and 
summed up. e sum was then converted into waiting time 
and divided by 0.5. For journeys in which the waiting time was 
longer than 4 min, it was assumed that passengers checked the 
bus schedule before the trip and, therefore, an average waiting 
time of 4 min was determined for these 16,771 trips. Accord-
ing to the initial estimate, the average waiting time on these 
trips was 6.1 min.
2.3 Calculating the travel time and speed
Because the calculations and denitions of travel speed vary 
signicantly, to ensure better comparability with research to 
date, the travel speed was calculated in four dierent ways, 
taking into account dierent distances and travel times, as 
shown in Table 1.
2.4 Walking and cycling speed
Cycling and walking travel times were modelled in OpenTrip-
Planner  (Morgan et  al,  2019). using the transport network 
created from the OpenStreetMap database  (OpenStreetMap 
contributors, 2015). ese data are of suciently high quali-
ty for Ljubljana to obtain suciently accurate results. In the 
OpenTripPlanner program, the default speed and weighting 
settings for individual road categories were used. e cycling 
speed was set at  17.7  km/h. Various estimates of the aver-
age speed of urban cyclists are used in the literature, rang-
ing from  15 to  19  km/h  (Ellison  & Greaves,  2011; Anders-
en, 2014; Kager et al., 2016). Because no data are available on 
the average speed of cyclists in Ljubljana, it is assumed that 
the above speed estimate is suitable. e walking speed was set 
at 4.8 km/h. Calculations were made for all origin-destination 
pairs. For walking and cycling, too, another 400 m were added 
to the distance between stops to calculate the eective travel 
speed, which added up to  1 min  30  s for cycling and  5  min 
for walking. We added another two minutes for cycling, as the 
time required to lock and unlock the bicycle.
2.5 Data merging and quality analysis
Aer conducting individual analyses, the data were merged 
into a joint database, in which the data analysed is collected 
for every journey. Journeys for which it was assumed that there 
were errors in the calculations were deleted from the database. 
It turned out that the criterion for merging trips into journeys 
that allows for less than 60 min for the transfer and a distance 
of less than  800  m between stops was insuciently accurate. 
us, to control for the quality of data, the coecient and 
dierence between lLPP line distance and lshortest were calculated. 
Where the lLPP line distance was signicantly greater then lshortest, 
this indicated that a transfer was wrongfully assigned instead 
of two separate journeys. us, all journeys in which lLPP line 
distance /lshortest < 0.8 or > 4 and lLPP line distance − lshortest < −100 m 
or > 100 m were eliminated from the database. Additionally, 
Calculating the speed of city bus trips: The case of Ljubljana, Slovenia
Urbani izziv, volume 31, no. 1, 2020
116
journeys were eliminated in which the actual travel speed was 
lower than 5 km/h or higher than 50 km/h. is way, errors 
were eliminated that might have occurred due to mistakes in 
the bus schedule or mistakes in merging individual trips into 
a journey where the waiting times were too long. In this situ-
ation, in reality a passenger can perform other activities in the 
meantime, such as go to a bar or shop, and then continue their 
journey. Such journeys are irrelevant in terms of studying travel 
speeds. Aer eliminating these inadequate ones,  70,768  trips 
remained out of the initial  74,085, based on which further 
analyses were performed.
3 Results
Based on the data analysed it is possible to conduct a series of 
analyses. Because the main purpose of this article is to analyse 
the travel speeds, the main results of analyses related to travel 
speed are presented below: rst, the results of the city bus 
analyses, followed by a comparison with walking and cycling 
travel speeds.
3.1 City bus
e main ndings of the city public transport analysis are pre-
sented in Table  2. Detailed information is presented in the 
subsections.
Table 2: Key results of the city public transport analysis.
Indicator Value
Effective total travel speed 10. km/h
Average actual distance travelled 4.8 km
Average effective distance travelled 4.1 km
Average waiting time 2.9 min
3.1.1 Average waiting time
One of the factors aecting the eective travel speed is the 
average time of waiting for the bus to arrive on the rst trip 
in the journey. e average waiting time is 2.9 min (SD = 1). 
Figure 1 shows the average waiting times and the share of wait-
ing in the total travel time, depending on the length of the 
journey. With longer journeys, on average passengers had to 
wait longer for the bus to arrive. One of the reasons for this 
is also that longer journeys had to start outside the city centre, 
where bus arrivals are less frequent. e longer the journey, the 
smaller the share of time spent waiting compared to the time 
spent for the entire journey.
3.1.2 Transfers
Users generally dislike transfers. e LPP network originat-
ed at a time when tickets were paid each time the passenger 
boarded the bus and hence one of the goals in designing the 
network was to reduce the need for transfers (Koblar, 2017). 
Table 1: Method of calculating the travel speed.
Presumed distance Presumed travel time
Effective total travel speed Effective distance travelled: lshortest + lwalking Total travel time: ttrip + twaiting + twalking
Total travel speed Distance travelled: lLPP line distance + lwalking Total travel time: ttrip + twaiting + twalking
Effective travel speed Effective distance travelled: lshortest Travel time: ttrip
Actual travel speed Actual distance travelled: lLPP line distance Travel time: ttrip
Whereby: 
lshortest: the shortest distance between the first and last stops calculated as the walking distance along pedestrian routes
lwalking: 400 m distance – the total walking distance to the first stop and from the last stop to the destination
lLPP line distance: distance travelled by bus; in the event of a transfer, the walking distance between the two transfer stops is taken into account
ttrip: time between boarding the bus on the first trip and alighting from the bus on the last trip of the journey; it also includes the time of 
transferring to the next line
twaiting: average time of waiting for the bus to arrive on the first trip in a journey
twalking: 5 min – the time required to walk 400 m, which is added as lwalking. This is an estimate based on how much time people are willing to 
spend walking to a bus stop (Tiran et al., 2019).
S. KOBLAR, L. MLADENOVIČ
Urbani izziv, volume 31, no. 1, 2020
117
In more developed networks, transfers are conceived as an 
important part of travel routes because they provide a com-
bination of various operators and systems and hence better 
public transport coverage (Mees, 2010; Dodson et al., 2011). 
In addition to 70,768 journeys, for which other analyses were 
also performed, the transfer analysis also included 17,614 user 
journeys that only made one trip on the day studied and hence 
their trips were unsuitable for calculating the alighting stops. 
Table  3 shows the number of journeys based on the number 
of transfers made.
3.1.3 Travel speeds
Travel speed is one of the factors that determine the quality 
of the public transport system. Table 4 shows the travel speeds 
based on the various criteria used and presented in Table 1.
In addition to the average speed, the distribution of the num-
ber of journeys shown in Figure 2 is also important. e his-
togram has a normal distribution shape, with slightly higher 
values on the right side.
Travel speed also depends on the length of the journey. In 
longer journeys, the waiting and walking times reduce the im-
pact on travel speed and so the speeds increase with the length 
of the journey. e eective travel speed curve is interesting: 
it is very high for short distances, resulting from the fact that 
the dierences between the distance travelled and the short-
est distance are smaller with shorter trips. In addition, these 
calculations do not account for the walking time to the bus 
stop and the waiting time.
3.2 Comparison with cycling and walking
To have a better idea of public transport travel speeds and 
to better understand the competitiveness of public transport 
over other forms of sustainable mobility, a comparison was 
also made with bicycle and walking travel speeds. In com-
Figure 1: Average waiting time and number of trips, depending on the length of the journey (author: Simon Koblar).
Table 3: Number of journeys based on the number of transfers made.
No. of transfers No. of journeys Share of all journeys (in %)
0 70,146 79.1
1 16,459 18.6
2 1,682 1.9
3 311 0.4
4 69 0.1
5 14 0.0
All journeys 88,681 100.0 %
Table 4: Calculated bus travel speeds.
Average speed 
(km/h)
SD (km/h)
Effective total travel speed 10.0 3.3
Total travel speed 11.3 3.4
Effective travel speed 15.7 6.2
Actual travel speed 17.6 5.7
Calculating the speed of city bus trips: The case of Ljubljana, Slovenia
Urbani izziv, volume 31, no. 1, 2020
118
paring the bus and bicycle travel speeds, eective total travel 
speeds were taken into account because they best reect the 
user experience. Eective total travel speeds increase with the 
length of the journey, due to a reduced impact of waiting and 
walking times for buses and a reduced eect of the additional 
time required to lock and unlock the bicycle. Bicycles are the 
fastest on all distances, with the dierence being the greatest 
in shorter journeys. On average, a bicycle would be  7.5 min 
faster than the bus. Only 8% of the journeys would have been 
faster with the bus and  46% of journeys would have been  5 
min faster with the bicycle.
Figure 2: Number of journeys by effective bus travel speed class (author: Simon Koblar).
Figure 3: Travel speed in correlation with the length of the journey (author: Simon Koblar).
S. KOBLAR, L. MLADENOVIČ
Urbani izziv, volume 31, no. 1, 2020
119
Due to the low speed of walking, only journeys up to 2 km were 
taken into account. On stretches shorter than 2 km, 926 jour-
neys (i.e., 7% of the total journeys shorter than 2 km) would 
have taken less time on foot than by bus. Also taking into 
account the journeys that are less than 1 minute faster by bus, 
the total number of these journeys adds up to 1,783 or 13%.
4 Discussion
is article presents new ndings related to the measurement 
of public transport quality and reveals great potential of the 
electronic payment system data for conducting further analy-
ses. Because analyses are performed based on the trips made, 
the results are especially interesting from the user perspective 
because they reect the user experience and provide insight 
into passenger behaviour. Because the payment system does 
not provide information on the alighting stop, determining the 
alighting stops represented a signicant challenge. To this end, 
available data were applied to a well-tested algorithm (Alsger 
et al., 2016), whereby the distance between stops was modelled 
in the GIS environment using pedestrian routes. is resulted 
in greater accuracy compared to the straight-line distance ap-
plied by Alsger et  al.  (2016). To determine the travel speeds 
the waiting time at the stop, the travel time, and the distance 
travelled were also calculated for each trip. e applied method 
for calculating the waiting time that also takes into account 
the time of day and relevant lines, yields more realistic results 
from the passenger perspective than the method of counting 
arrivals at peak times frequently used in other studies of public 
transport quality in Ljubljana (Bole, 2004; Tiran et al., 2015).
Because the shortest distance between the origin and destina-
tion is especially important from the passenger perspective, the 
shortest distance in the transport network was also modelled 
in addition to the distance travelled on a public transport line. 
Various methods are used to calculate the travel speed and 
hence four dierent methods were applied, using dierent 
distances and times. From the user perspective and compared 
to other travel modes, the most relevant is the eective total 
travel speed, which on average amounts to 10.0 km/h; this is 
signicantly lower than the average actual travel speed of 17.6 
km/h. Commercial speed is the only information that has been 
available for the entire network in comparable form to date. 
According to the LPP data, the commercial speed, which only 
takes into account the individual trip without transfers to other 
lines, is 18 km/h (Šmajdek, 2011), demonstrating the accura-
cy of the analyses conducted. e substantial dierences in 
results indicate the importance of selecting the travel speed 
calculation method.
By calculating the travel speeds, the rst hypothesis was also 
conrmed. Based on the electronic payment system data and 
bus schedules it is possible to determine the travel speed of 
bus trips. A comparison of bus travel speeds with walking and 
cycling showed that the bus is poorly competitive with bicy-
cles. On average, equivalent trips took 7.5 min longer by bus 
than by bicycle. is also conrmed the second hypothesis. An 
average bus trip would have taken less time if made by bicycle. 
Some shorter routes would even have been travelled faster on 
foot, which points to frequently irrational passenger decisions. 
Most of these shorter trips are made in the city centre, where 
buses are very full as it is. e ratio between the bus and cy-
Figure 4: Comparison of bus and bicycle speeds and travel times (author: Simon Koblar).
Calculating the speed of city bus trips: The case of Ljubljana, Slovenia
Urbani izziv, volume 31, no. 1, 2020
120
cling travel speeds is most likely one of the reasons for the 
increase in cycling (Klemenčič et al., 2014) and the decline in 
the number of bus passengers in recent years (Ljubljanski pot-
niški promet, 2019). In addition to travel speeds, insight was 
also provided into passenger behaviour in terms of transfers. It 
turned out that despite changes to the payment system, which 
allows free transfers within  90  min aer the rst validation, 
only 20.9% of journeys include transfers. is probably results 
from the network’s design, which is supposed to reduce the 
number of required transfers as much as possible, and partly 
also from the fact that  (predominantly elderly) users tend to 
only accept change and change their habits slowly.
e method applied also has certain deciencies and some 
could be eliminated through further research and more com-
plex methods. Due to the large quantity of the electronic pay-
ment system data, complete control over their quality cannot 
be guaranteed. Certain errors can already arise in determining 
the alighting stops, whereby additional parts of the trips for 
which there are not suitable data are eliminated. In terms of 
data quality, what is especially problematic is merging several 
trips into a journey, which could be improved through more 
complex methods  (Assemi et  al.,  2020). e key step in this 
study was the elimination of outliers from further calculations. 
Unfortunately, the accuracy of the alighting stops determined 
cannot be estimated, which, modelling on Wang et al. (2011), 
could have been done through eld research and by comparing 
the results. In addition, using dierent presumptions about the 
random passenger arrival at the bus stop would have yielded 
somewhat dierent results in calculating the average waiting 
time  (Amin-Naseri  & Baradaran,  2015). In determining the 
walking distance, a uniform value of 400 m was used because 
no data are available on what distance the users of the Ljublja-
na public transport system actually walk. e cycling speed 
applied in the study was a mere estimate, too. Due to many 
elements that aect it  (e.g.,  the quality of the cycling infra-
structure, waiting at trac lights, and ultimately the type of 
cyclist and bicycle used), the results could have been dierent 
if a dierent assumed speed have been used. By improving the 
quality of the cycling infrastructure and increasing the share 
of electric bicycles the average cycling speeds can be expected 
to rise. A certain degree of error also occurs in calculating 
the bus speed, which was determined based on the available 
bus schedules. e actual speeds always deviate from these, 
especially at the stops close to the end of the lines. A solution 
would be to use the data from the vehicle tracking system, 
based on which the bus speeds could be determined more 
accurately (Wang et al., 2011). 
e public transport payment system data also make it possi-
ble to conduct a series of other analyses (Pelletier et al., 2011; 
Ali et  al.,  2016; Trépanier  & Morency,  2017), which would 
be prudent in the future. Good familiarity with the public 
transport system and passenger behaviour may be of great help 
in introducing improvements to the system, which are vital 
for Ljubljana due to the poor competitiveness of its public 
transport and the inappropriate design of its network (Koblar 
et  al.,  2018). Specically, it is vital to reverse the decreasing 
trend in the number of passengers because only this way the 
targeted share of public transport trips can be achieved  (Mi-
lovanović, 2017), which would contribute to a reduced envi-
ronmental impact. On the other hand, improvements in the 
public transport system alone are not enough; a better integra-
tion of spatial and transport planning is also required  (Plev-
nik, 1997), which especially applies to the well-served public 
transport corridors (Šašek Divjak, 2004).
5 Conclusion
e method of analysing the public transport payment sys-
tem and measuring the travel speed presented and applied to 
Ljubljana is one of the few attempts to measure the quality 
of the public transport network based on trips actually made. 
e eective total travel speed reects the user experience sig-
nicantly better than the more widely used commercial speed 
measurements. In turn, comparing the bus trips to cycling and 
walking suitably contextualizes these speeds. Calculating the 
speeds also yielded other important information, such as the 
travel time, the distance travelled, the average waiting time, 
and the number of transfers. In the future, the actual distance 
walked to the bus stop should be taken into account, the bus 
speed should be calculated from the vehicle tracking system, 
and greater attention should be dedicated to quality control, 
especially in determining the alighting stops and merging trips 
into journeys. In addition, analysis should cover a longer pe-
riod. e method applied is very useful for monitoring the 
use of the public transport system and improving it, which 
could reverse the falling trend in the number of passengers. e 
current ndings for Ljubljana alone can be used by transport 
planners and LPP to introduce changes that would increase 
the competitiveness of the public transport systems.
Simon Koblar
Urban Planning Institute of the Republic of Slovenia, Ljubljana, 
Slovenia
E-mail: simon.koblar@uirs.si
Luka Mladenovič
Urban Planning Institute of the Republic of Slovenia, Ljubljana, 
Slovenia
E-mail: luka.mladenovic@uirs.si
S. KOBLAR, L. MLADENOVIČ
Urbani izziv, volume 31, no. 1, 2020
121Calculating the speed of city bus trips: The case of Ljubljana, Slovenia
Acknowledgments
The study was financially supported through the projects ASTUS 
co-funded via the Alpine Space programme, and DG MOVE – SUMI, 
as part of which part of the analyses were conducted. Special thanks 
go to LPP for providing the relevant data and Matej Žebovec for 
providing assistance in analysing the alighting stops.
References
Ali, A., Kim, J. & Lee, S. (2016) Travel behavior analysis using smart 
card data. KSCE Journal of Civil Engineering, 20(4), pp. 1532–1539. 
DOI: 10.1007/s12205-015-1694-0
Alsger, A., Assemi, B., Mesbah, M. & Ferreira, L. (2016) Validating and im-
proving public transport origin–destination estimation algorithm using 
smart card fare data. Transportation Research Part C: Emerging Technolo-
gies, 68, pp. 490–506. DOI: 10.1016/j.trc.2016.05.004
Amin-Naseri, M. R. & Baradaran, V. (2015) Accurate estimation of aver-
age waiting time in public transportation systems. Transportation Sci-
ence,  49(2), pp. 213–222. DOI: 10.1287/trsc.2013.0514
Andersen, M. C. (2014) Bikes beat metro in Copenhagen. Available at: 
http://www.copenhagenize.com/2014/04/bikes-beat-metro-in-copenha-
gen.html (accessed 30 Jan. 2020).
Assemi, B., Alsger, A., Moghaddam, M., Hickman, M. & Mesbah, 
M. (2020) Improving alighting stop inference accuracy in the trip chain-
ing method using neural networks. Public Transport, 12(1), pp. 89–121. 
DOI: 10.1007/s12469-019-00218-9
Bole, D. (2004) Geografija javnega potniškega prometa na primeru 
Ljubljane. Geografski vestnik, 76(2), pp. 21–32.
Carreira, R., Patrício, L., Natal Jorge, R. & Magee, C. (2014) Understand-
ing the travel experience and its impact on attitudes, emotions and 
loyalty towards the transportation provider – A quantitative study with 
mid-distance bus trips. Transport Policy, 31, pp. 35–46.  
DOI: 10.1016/j.tranpol.2013.11.006
Čelan, M. & Lep, M. (2020) Bus-arrival time prediction using bus net-
work data model and time periods. Future Generation Computer Sys-
tems, 110, pp. 364–371. DOI: 10.1016/j.future.2018.04.077
Celcer, E. (2009) Merila za oceno kakovosti poteka prog javnega prometa. 
Bachelor’s thesis. Univerza v Ljubljani, Fakulteta za gradbeništvo in 
geodezijo.
Constantinescu, A., Popa Mitroi, G., Trotea, M. & Bogdan, M. L. (2018) 
Analysis of the commercial speed of a public passenger transport line 
from Craiova-Romania. Applied Mechanics and Materials, 880, pp. 389–
396. DOI: 10.4028/www.scientific.net/AMM.880.389
Cui, A. (2006) Bus passenger origin-destination matrix estimation using 
automated data collection systems. Doctoral dissertation. Boston, Mas-
sachusetts Institute of Technology.
Dodson, J., Mees, P., Stone, J. & Burke, M. (2011) The principles of public 
transport network planning: A review of the emerging literature with select 
examples. Brisbane, Urban Research Program, Griffith University.
Ellison, R. B. & Greaves, S. (2011) Travel time competitiveness of cycling 
in Sydney, Australia. Transportation Research Record: Journal of the Trans-
portation Research Board, 2247(1), pp. 99–108. DOI: 10.3141/2247-12
Farzin, J. (2008) Constructing an automated bus origin-destination 
matrix using farecard and global positioning system data in Sao Paulo, 
Brazil. Transportation Research Record: Journal of the Transportation 
Research Board, 2072(1), pp. 30–37. DOI: 10.3141/2072-04
Gabrovec, M. & Bole, D. (2006) Dostopnost do avtobusnih postajališč. 
Geografski vestnik, 78(2), pp. 39–51.
Gabrovec, M. & Razpotnik Visković, N. (2012) Ustreznost omrežja javne-
ga potniškega prometa v Ljubljanski urbani regiji z vidika razpršenosti 
poselitve. Geografski vestnik, 84(2), pp. 63–72.
Gabrovec, M. & Razpotnik Visković, N. (2018) Dostopnost do javnega 
potniškega prometa kot pogoj za socialno vključenost dijakov. Geograf-
ski vestnik, 90(2), pp. 109–120. DOI: 10.3986/gv90206
Gojčič, M. (ed.) (2018) Celostna prometna strategija Ljubljanske urbane 
regije: za ljudi in prostor v inovativni in napredni regiji. Ljubljana, Region-
alna razvojna agencija Ljubljanske urbane regije.
Jang, W. (2010) Travel time and transfer analysis using transit smart 
card data. Transportation Research Record: Journal of the Transportation 
Research Board, 2144(1), pp. 142–149. DOI: 10.3141/2144-16
Kager, R., Bertolini, L. & Te Brömmelstroet, M. (2016) Characterisation 
of and reflections on the synergy of bicycles and public transport. 
Transportation Research Part A: Policy and Practice, 85, pp. 208–219. 
DOI: 10.1016/j.tra.2016.01.015
KFH Group (2013) Transit capacity and quality of service manual, 3rd edi-
tion, TCRP report. Washington, D.C, Transportation Research Board. 
DOI: 10.17226/24766
Klemenčič, M., Lep, M., Mesarec, B. & Žnuderl, B. (2014) Potovalne 
navade prebivalcev v mestni občini Ljubljana in Ljubljanski urbani regiji. 
Ljubljana, Mestna občina Ljubljana.
Koblar, S. (2017) Predlog alternativnega omrežja javnega potniškega 
prometa v Ljubljanski urbani regiji. Master’s thesis. Ljubljana, Univerza v 
Ljubljani, Filozofska fakulteta.
Koblar, S. & Mladenovič, L. (2020): Comparison of travel speeds be-
tween different sustainable modes in Ljubljana. In: Ceste 2020. Zbornik 
radova, pp. 198–210. Rovinj, Tom Signal.
Koblar, S., Plevnik, A. & Repe, B. (2018) Predlog alternativnega omrežja 
javnega potniškega prometa v Ljubljanski urbani regiji. In: Žnidarič, 
A. (ed.) Slovenski kongres o prometu in prometni infrastrukturi, 14, 
pp. 1–8. Ljubljana, DRC – Družba za raziskave v cestni in prometni 
stroki Slovenije.
Koblar, S. & Žebovec, M. (2018) Analiza potovalnih vzorcev uporabnikov 
mestnega potniškega prometa v Ljubljani. In: Ciglič, R. (ed.) GIS v Slo-
veniji 14: Pokrajina v visoki ločljivosti, pp. 165–173. Ljubljana, Založba 
ZRC.
Koren, M. (2016) Načrtovanje in optimiranje linij javnega potniškega 
prometa. Bachelor’s thesis. Ljubljana, Univerza v Ljubljani, Fakulteta za 
gradbeništvo in geodezijo.
Kozina, J. (2010) Modeliranje prostorske dostopnosti do postajališč 
javnega potniškega prometa v Ljubljani. Geografski vestnik, 82, pp. 97–
107. DOI: 10.3986/gv90206
Li, D., Lin, Y., Zhao, X., Song, H. & Zou, N. (2011) Estimating a transit 
passenger trip origin-destination matrix using automatic fare collection 
system. In: Xu, J., Yu, G., Zhou, S. & Unland, R. (eds) Database systems 
for adanced applications. DASFAA 2011. Lecture Notes in Computer 
Science, pp. 502–513. New York, Springer.  
DOI: 10.1007/978-3-642-20244-5_48
Ljubljanski potniški promet (2019) Letno poročilo 2018. Ljubljana.
Loong, C. & El-Geneidy, A. (2016) It’s a matter of time: assessment of 
additional time budgeted for commuting to McGill university across 
modes. Transportation Research Record: Journal of the Transportation 
Research Board, 2565(1), pp. 94–102. DOI: 10.3141/2565-11
Urbani izziv, volume 31, no. 1, 2020
122
Lu, D. (2008) Route level bus transit passenger origin-destination flow es-
timation using APC data: Numerical and empirical investigations. Master’s 
thesis. Ohio, Ohio State University.
Mees, P. (2010) Transport for suburbia: Beyond the automobile age. Lon-
don, Sterling, VA, Earthscan.
Milovanović, K. (ed.) (2017) Celostna prometna strategija Mestne občine 
Ljubljana. Ljubljana, Mestna občina Ljubljana.
Morgan, M., Young, M., Lovelace, R. & Hama, L. (2019) OpenTripPlanner 
for R. Journal of Open Source Software, 4(44), 1926.  
DOI: 10.21105/joss.01926.
Morris, E. A. & Guerra, E. (2015) Mood and mode: does how we travel 
affect how we feel? Transportation, 42(1), pp. 25–43.  
DOI: 10.1007/s11116-014-9521-x
Mosallanejad, M., Somenahalli, S. & Mills, D. (2019) Origin-destination 
estimation of bus users by smart card data. In: Geertman, S., Zhan, 
Q., Allan, A. & Pettit, C. (eds.) Computational urban planning and 
management for smart cities. Lecture notes in geoinformation and 
cartography, pp. 305–320. Cham, Springer International Publishing. 
DOI: 10.1007/978-3-030-19424-6_17
Munizaga, M., Nuñez, C. & Gschwender, A. (2016) Smart card data 
for wider transport system evaluation. In: Kurauchi, F. & Schmok-
er, J. -D.  (eds.) Public transport planning with smart card data, 
pp. 163–179. Boca Raton, FL, CRC Press, Taylor & Francis Group. 
DOI: 10.1201/9781315370408-13
OpenStreetMap contributors (2015) Planet dump [Data file from $date 
of database dump$]. Available at: https://planet.openstreetmap.org (ac-
cessed 7 Mar. 2020).
Pelletier, M.-P., Trépanier, M. & Morency, C. (2011) Smart card data use 
in public transit: A literature review. Transportation Research Part C: 
Emerging Technologies, 19(4), pp. 557–568.  
DOI: 10.1016/j.trc.2010.12.003
Plevnik, A. (1997) The importance of integrating urban and traffic plan-
ning. Urbani izziv, 32–33, pp. 141–146.  
DOI: 10.5379/urbani-izziv-en-1997-32-33-006
Schmöcker, J.-D., Kurauchi, F., in Shimamoto, H. (2016): An overview on 
opportunities and challenges of smart card data analysis. In: Kurauchi, 
F., & Schmöcker, J.-D. (ed.) Public transport planning with smart card 
data, str. 1–11. Boca Raton, FL, CRC Press, Taylor & Francis Group. 
DOI: 10.1201/9781315370408-2
Statistični urad RS (2017) Razlogi prebivalcev Slovenije (15–84 let), da 
niso dnevno uporabljali javna prevozna sredstva, po spolu (v %), Slovenija, 
jeseni 2017. Ljubljana.
Stradling, S., Carreno, M., Rye, T. & Noble, A. (2007) Passenger percep-
tions and the ideal urban bus journey experience. Transport Poli-
cy, 14(4), pp. 283–292. DOI: 10.1016/j.tranpol.2007.02.003
Šabić, A. (2015) Učinki spremembe prometnega režima na Slovenski cesti 
s posebnim poudarkom na javnem potniškem prometu. Bachelor’s thesis. 
Ljubljana, Univerza v Ljubljani, Fakulteta za gradbeništvo in geodezijo.
Šašek Divjak, M. (2004) Transport corridors and settlements in a region: 
Linking settlements to public transport. Urbani izziv, 15(1), pp. 97–101. 
DOI: 10.5379/urbani-izziv-en-2004-15-01-003
Šmajdek, J. (2011) Uvajanje sodobnega plačilnega sistema v javnem mest-
nem potniškem prometu. Certificate thesis. Koper, Univerza v Ljubljani, 
Fakulteta za pomorstvo in promet.
Tiran, J., Lakner, M. & Drobne, S. (2019) Modelling walking accessibil-
ity: A case study of Ljubljana, Slovenia. Moravian Geographical Re-
ports, 27(4), pp. 194–206. DOI: 10.2478/mgr-2019-0015
Tiran, J., Mladenovič, L. & Koblar, S. (2015) Dostopnost do javnega 
potniškega prometa v Ljubljani po metodi PTAL. Geodetski vest-
nik, 59(4), pp. 723–735.
Trépanier, M. & Morency, C. (2016) Evaluation of bus service key 
performance indicators using smartcard data. In: Kurauchi, F. & 
Schmoker, J.-D. (eds.) Public transport planning with smart card 
data, pp. 181–196. London, CRC Press, Taylor & Francis Group. 
DOI: 10.1201/9781315370408-14
Trépanier, M., Tranchant, N. & Chapleau, R. (2007) Individual trip des-
tination estimation in a transit smart card automated fare collection 
system. Journal of Intelligent Transportation Systems, 11(1), pp. 1–14. 
DOI: 10.1080/15472450601122256
Vanhanen, K. & Kurri, J. (2007) Quality factors in public transport. Sum-
mary report of the Public Transport Research Programme JOTU 2004–
2007. Helsinki, Ministry of Transport and Communication.
Wang, W. (2010) Bus passenger origin-destination estimation and travel 
behavior using automated data collection systems in London, UK. Doctoral 
dissertation. Boston, Massachusetts Institute of Technology.
Wang, W., Attanucci, J. P. & Wilson, N. H. (2011) Bus passenger origin-
destination estimation and related analyses using automated data 
collection systems. Journal of Public Transportation, 14(4), pp. 131–150. 
DOI: 10.5038/2375-0901.14.4.7
Yan, F., Yang, C. & Ukkusuri, S. V. (2019) Alighting stop determination 
using two-step algorithms in bus transit systems. Transportmetrica A: 
Transport Science, 15(2), pp. 1522–1542.  
DOI: 10.1080/23249935.2019.1615578
Zhao, J., Rahbee, A. & Wilson, N. H. (2007) Estimating a rail passenger 
trip origin-destination matrix using automatic data collection systems. 
Computer-Aided Civil and Infrastructure Engineering, 22(5), pp. 376–387. 
DOI: 10.1111/j.1467-8667.2007.00494.x
S. KOBLAR, L. MLADENOVIČ