measuring tourist activities in cities using geotagged photography
TRANSCRIPT
This article was downloaded by: [Thammasat University Libraries]On: 09 October 2014, At: 04:40Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Tourism Geographies: An InternationalJournal of Tourism Space, Place andEnvironmentPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/rtxg20
Measuring tourist activities in citiesusing geotagged photographyBálint Kádára
a Department of Urban Planning and Design, Budapest Universityof Technology and Economics, Budapest, HungaryPublished online: 06 May 2014.
To cite this article: Bálint Kádár (2014) Measuring tourist activities in cities using geotaggedphotography, Tourism Geographies: An International Journal of Tourism Space, Place andEnvironment, 16:1, 88-104, DOI: 10.1080/14616688.2013.868029
To link to this article: http://dx.doi.org/10.1080/14616688.2013.868029
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.
This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Measuring tourist activities in cities using geotagged photography
B�alint K�ad�ar*
Department of Urban Planning and Design, Budapest University of Technology and Economics,Budapest, Hungary
(Received 26 February 2012; accepted 25 August 2013)
Established methods to record tourist activities in urban centres cannot producequantitative data with precise spatial references. Analysing geographically positionedphotography retrieved from image hosting web services combines the accuracy levelsof GPS tracking and the quantitative advantages of the large accessible data-sets ofinternet communities. In this paper the correlation of data-sets obtained from Flickr.com with statistical data, morphological constrains and the attributes of attractions aretested. Correlation between registered bed nights and geotagged tourist photographynumbers was calculated analysing 16 European cities. Three tourist-historic citieswith similar tourist markets were compared more deeply. The spatial patterns oftourist activity in Vienna, Prague and Budapest showed many similarities and somerelevant differences rooted in the morphological constraints and the different level oflocal and tourist activities at specific sites.
Keywords: tourist tracking; Flickr; tourist photography; Vienna; Prague; Budapest
Introduction
Measuring visitors’ space usage in the tourist-historic city
Tourists in cities move between sights, monuments, museums or cultural events and pla-
ces of shopping, dining and interaction with other people. These attractions are all the ele-
ments that motivate a travel, are already marked for the visitor through various media,
and are being consumed during the visit even if no active consumption is possible, at least
by taking a photograph. While the singular points of attraction are well definable even in
the urban context, the tracking of tourists is a difficult task, especially in a tourist-histori-
cal city (Ashworth & Turnbridge, 1990), where the majority of attractions, monuments
and events are embedded in historical urban layouts with an evolved and protected mor-
phology, used also by the local community with their own infrastructures, businesses and
cultural uses. Such multifunctionality imposed limitations for researchers. It was already
argued by Stansfield that urban tourism is not quantifiable, therefore ignored by research-
ers who could immerse themselves in the more measurable rural tourism (Stansfield,
1964). In the past decades there was much progress in this field. A never seen quantity of
case studies appeared, motivations and global trends revealed, the supply side of urban
tourism was mapped, but the demand side is still not fully understood on behalf of miss-
ing data (Pearce, 1998). Pearce (2001) noted how different fields researching urban tour-
ism are missing an integrative framework. Ashworth and Page (2011), gathering the most
recent progress in this field, also pointed out the vagueness of urban tourism research
studies consisting of individual, unconnected individual case studies. Still, a more
*Email: [email protected]
� 2014 Taylor & Francis
Tourism Geographies, 2014
Vol. 16, No. 1, 88–104, http://dx.doi.org/10.1080/14616688.2013.868029
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
efficient understanding of tourist space usage is needed, also because this field is rela-
tively neglected among the others as they also note: ‘Given the quantitative importance
of urban tourism, it is curious that very little attention has been given to questions about
how tourists actually use cities.’ (Ashworth & Page, 2011). To face this question, a better
mapping of local and tourist space usage and the conceptualization of tourist movements
in cities are needed.
This paper proposes a quantitative method capable to complement other attempts to
define how tourists move in the urban context and how do they consume its spaces. The
intent is to give a better geographical base for comparative studies of urban tourism, using
spatial data-sets available for all main urban destinations: photography uploaded and geo-
tagged in image sharing sites.
Comparative, or ‘multiple city studies’ are rare in the field of urban tourism, and
they ‘. . .tend to be more narrowly focused and examine a specific aspect of a broader
problem. . .’ (Pearce, 2001). Beside the few examples collected by Pearce some studies
comparing Hong Kong and Singapore (Henderson, 2002) and works by Van der Borg
comparing cultural tourism in some European cities are relevant (Van der Borg, Costa,
& Gotti, 1996; Russo & Van der Borg, 2002). These either compare arrival and hotel
numbers, aspects of the demand side of tourism, or just have a focus on the impact to
residents’ attitudes. Comparing the geographical aspects of tourism in different urban
destinations could lead to a better understanding of issues like congestion, conflicts
with locals and other sociological and economical aspects, understanding better the
correlations between urban design, tourism planning, sociological and economical
aspects.
Few studies achieved to refine the basic model of urban tourism setup by Ashworth
and Turnbridge (1990) in ‘the tourist-historic city’. Pearce (1998) could map the structure
and functions of some important tourist districts in Paris, analysing the supply side of
tourism. He explored the limits of this method, mapping which services are being used by
locals and which by tourists, and how these work in different urban contexts and scales,
including the microscale of the singular attraction (Pearce, 1999). In his papers, he
stresses the need to understand also the supply side, the behaviour and movement of tou-
rists in the urban space.
An important paper trying to conceptualize the movement of tourists in urban destina-
tions is the one of Lew and McKercher (2006), further developed in a study using data
from surveys and time-space diaries filled out by independent tourists in Hong Kong,
identifying 11 movement styles (McKercher & Lau, 2008). Similar methods of question-
naires were used by Shoval and Raveh (2004) to cluster the attractions in Jerusalem
according to the different characteristics of tourists, and by Hayllar and Griffin (2005),
who could define the most important themes in tourist experiences related to the physical
environment and atmosphere of The Rocks district in Sydney. All these studies faced the
restrictions of the survey method: the limited geographical preciseness and the small
numbers of agents involved.
Two new technologies helped to get over these limitations: the use of global posi-
tioning systems (GPS) and the analysis of mobile phone cell information (Shoval &
Isaacson, 2007). GPS trackers are reliable and accessible devices, used by many scholars
since the opening of the navigation system for civilian use in 2000 (Shoval & Isaacson,
2010). Pettersson and Zillinger (2011) combined GPS tracking and questionnaires to
gather information on tourists’ experiences in relationship of their movements at a sports
event. Modsching, Kramer, ten Hagen, and Gretzel (2008) used GPS tracking data to
trace the activity areas of tourists, drawing the most visited hubs and paths in the
Tourism Geographies 89
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
German city of G€orlitz. Shoval (2008) used three-dimensional (3D) bar diagram repre-
sentations of tourist activities on the map showing the intensity of visitors’ activity in
the Old City of Akko. Shoval, McKercher, Ng, and Birenboim (2011) tracked the day
trips of tourists based in different hotels of Hong Kong, and compared the differences
of spatiotemporal patterns in first and repeat visitors (McKercher, Shoval, Ng, and
Birenboim, 2012). The quantitative dimension of these visual representations is rather lim-
ited: 557 tourists supplied the data for the research in Hong Kong, while only 134 tracks
were used in the case of Akko. In this quantitative aspect the usage of mobile phone cell
information to retrieve geographical data is more purposive. Today most people use their
devices even during vacations abroad. Researchers do not have to make field work to
gather such data as the mobile phone operators record all activity related to the devices
using their networks. Ahas, Aasa, Mark, Pae, and Kull (2007) showed the potentials in this
method, analysing the activity of mobile phones in Estonia for a time span of an entire
year. They could retrieve data regarding the time, geographical position and nationality of
all activity from foreign mobile phones using the network of a specific carrier. The results
were controlled in correlation to the data-sets of accommodation available nationwide.
This research showed the usability of such data at a regional level, but exact geographical
positioning of cellular phone activity is limited by the cell layouts of the network. In an
urban centre cellular positioning can be 100 meters accurate, but this value is still not ade-
quate to measure tourist activity between the attractions of a dense urban situation (Shoval
& Isaacson, 2010). The ideal tourist tracking method should combine the immense quantity
of data peculiar to the mobile cell information methodology, and the accuracy of the GPS
services. In the future with the spreading of smart phones this combination could appear
extensively, but until then other methods would be needed.
Using travel photography to measure tourist activity
Most tourists take photographs, creating records of their visit a final proof that they did
consume the experience given by the city (Urry, 1990). The role of photography in the
tourist experience had been analysed after Urry by many scholars (Garrod, 2009; Jenkins,
2003; Larsen, 2006). The subjects of tourist photography describe well a destination,
revealing much of its tourist offer and demand. The geographic positioning of each photo
in itself carry important information. With digital imaging and geotagged storage, a rele-
vant number of photographs with precise geographical data is uploaded and shared on
social websites like Panoramio.com or Flickr.com. The motivation of tourists to record
and share consumed personal experiences by their own photography meets well the possi-
bilities offered by these online services. Some cameras, including the growing number of
smart phones have built-in GPS antennas, allowing the automation of geotagging, while
the manual placement of singular images on a map is also possible on these sites. The
most popular website Flickr.com (http://www.Flickr.com) stores 200 million geotagged
images today. The application programming interface (API) of Flickr.com makes possible
the analysis of all of these images (Popescu & Grefenstette, 2009), a database large
enough to make relevant statistical deductions from their data, even though only a portion
of all photographers upload such geotagged images. A survey made among 1466 house-
holds of Hong Kong focused on tourist photography and the use of social media to share
it (Lo, McKercher, Lo, Cheung, & Law, 2011). The study showed that 89% of those trav-
eling overnight for tourism took photographs. Out of this group, 41.4% posted some of
these pictures online and 16.5% on sites like Flickr.com. The study also pointed out a
strong relationship between the age and education and the willingness to use such tools.
90 B. K�ad�ar
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
Younger and more educated people use these services more often, but all age groups are
well represented, much more than in other social media services.
Many emerging studies rely on data mining the open API of Flickr.com. Some of
these researches analyse the content and the given verbal tags of these images, relating
them with the geographic position (Hollenstein & Purves, 2010; Pang et al., 2011). Others
explore the viewpoints of singular attractions, reconstructing visual information of land-
marks (Crandall & Snavely, 2012; Ji, Gao, Zhong, Yao, & Tian, 2011). Very few attempts
have been made to retrieve quantitative data of tourist space usage from these databases.
The most promising study analysed photographs uploaded to Flickr.com in the Province
of Florence, Italy between 2005 and 2007 (Girardin, Fiore, Ratti, & Blat, 2008). The
study presented various geovisualizations to show tourist concentration and spatiotempo-
ral flows. These visualizations did show the main points of tourist activity, but the
selected method of heat maps could only provide generic information. The most interest-
ing information was visible only on a regional level: the authors traced the movements of
the individual users, foreigners and Italians separately, following the sequence of their
photographs throughout the province and all Italy. More precise maps were created based
on the same principle of separating tourists and locals by Gede (2012), further developed
by K�ad�ar and Gede (2013). These geovisualizations were optimized to be comparable
and measurable at an urban level, therefore a mesh of 10 by 10 meters was used with
each cell being a bar of a diagram positioned on Google Earth in a similar way. Shoval
(2008) represented the time-space diagrams obtained from GPS trackers of tourists in
Akko. This combination of the accuracy of GPS devices and the immense data-sets of
social image sharing sites can add new answers to the original question about how tourists
use urban spaces. This paper uses the method of K�ad�ar and Gede (2013) to measure tour-
ist activities in three Central-Eastern European capital cities: Vienna, Prague and Buda-
pest. These three destinations are tourist-historic cities, complex urban metropolises,
visited by the same type of tourists, often one after the other (Puczk�o & R�atz, 2000). Thepaper aims to demonstrate the validity of this method in finding space usage patterns,
measuring the so far immeasurable attendance to sites and spaces of tourist interest, and
verifying some behavioural differences between tourists and local visitors.
Methods
Retrieving data from geotagged photography uploaded to Flickr
Using the method of K�ad�ar and Gede (2013), databases for Vienna, Prague and Budapest
were created from the geographical references of all photographs available on Flickr.com
positioned in the urban areas of these cities. The data were downloaded on 21 January
2013, therefore all photographs taken before 2013 are part of the data-set. The data down-
loaded were the following: geographical coordinates, user ID, date of photograph from
the EXIF of the image and the accuracy level of geographic positioning. To discard false
or inaccurate positioning data, only photographs with Flickr’s maximum accuracy level
of 16 were taken into consideration in this research. This means that either a GPS enabled
device was needed to geotag these photographs or the user must have placed the image
on the map interface at the maximum street-level zoom.
To define patterns in urban tourist space usage and to analyse the different photo-
graphing behaviour of tourists and locals, a separation was needed between the photo-
graphs of these two groups. Girardin et al. (2008) used a 30-day period limitation in their
algorithm to separate tourists and locals, which is a valid assumption for their study
Tourism Geographies 91
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
analysing vacation patterns at a regional level. In urban tourism much shorter periods are
proper. The three cities analysed have an average length of stay between 2 and 3 days
(Puczk�o & R�atz, 2000). In this paper photographs taken from users active in the area for a
maximum of five consecutive days are considered tourist images. It was assumed and
later proved that those on longer stays or repeated visitors are interested also in less tour-
isty sites, and usually take more photographs in one place, just like locals do.
For all numerical analysis, only data for the full year of 2011 were taken into consid-
eration, both for photographs geotagged and for official statistics of tourist arrivals, bed
nights and singular attraction attendance, obtained from TourMIS (2012). Other data con-
cerning tourism like expenditure, nationalities or supply of services did not result useful,
therefore these were not analysed. The downloaded data-sets were visualized and ana-
lysed by the methods of K�ad�ar and Gede (2013). The sightly 3D bar maps were created
positioning both tourist and local activity on Google Earth, while a counter tool simplified
the process of counting photographs in selected geographical areas. Within the area
selected not only the number of geotagged photographs were counted – separated to local
and tourist groups, but also the number of local and tourist individuals uploading any
number of photographs taken in that area, resulting in the average number of photographs
a user takes in a site.
The potentials of this method lie in the combination of geographically accurate GPS
tracking and large databases of online image sharing, with the possibilities to separate
data of users considered tourists and locals. To verify the validity of the method, available
tourism statistics data-sets and the number of tourist photographs geotagged were first
compared. After this the assumption that the places where these images have been taken
correlate with their actual tourist use was analysed. At the urban level the patterns of tour-
ist photography distribution were compared with the morphology of the historical centres
of Vienna, Prague and Budapest. At the level of singular attractions, questions regarding
the correlation between actual visitors and numbers of photographs and photographers,
differences between patterns of tourist and local photography and questions regarding the
complexity and popularity of a photographed tourist site were analysed.
Results
Correlations and comparisons at the urban level
The basic validity of this method was first proved comparing the numbers of uploaded
geotagged photography with city tourism statistics. Flickr data from 16 European tourist-
historic cities were downloaded and four correlation coefficients between the number of
photographs, users, registered bed nights spent and arrival numbers were calculated
(Table 1). The correlation coefficient between registered tourist bed nights and both user
numbers and number of photographs is a considerable 0.92. In a logical way the correla-
tion with registered arrival numbers is lower (0.89): bed night statistics relate with the
total time tourists had spent in a city, and the same is true for the number of photographs
taken in a city in the same period. The proven correlation encouraged further work with
the method.
The patterns drawn from geotagged photography were compared with the urban spa-
tial structures and tourist offers of the three Central European capital cities. The geovisu-
alizations of tourist photography draw out visibly the main attractions and routes visited
by tourists (Figure 1), relieving intensities of tourist usage that no other methods could
visualize so far. These visible patterns correlate well to the historical structure and
92 B. K�ad�ar
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
Table1.
Arrivalandbed
nightstatistics,Flickrusersandthenumber
oftheirphotographsgeotagged
retrieved
for16Europeancities
from
theyear2011.
Year2011
Arrivals
Bed
nights
Geotagged
images
Flickr
users
Tourists’
images
Tourist
users
Locals’
images
Local
users
Population
Vienna
5,694,392
12,246,204
42,062
1698
10,163
924
31,899
774
1,731,236
Prague
5,132,042
13,214,304
34,978
1908
18,257
1273
16,721
635
1,262,106
Budapest
2,813,139
6,598,989
27,557
1646
9862
910
17,695
736
1,741,041
Berlin
9,866,088
22,359,470
102,987
4667
24,626
2409
78,361
2258
3,520,061
Barcelona
7,894,885
16,911,093
92,657
5145
27,384
2680
65,273
2465
3,218,071
Madrid
8,330,158
16,376,844
78,756
4678
13,800
2102
64,956
2576
3,265,038
M€ un
chen
5,931,052
11,738,112
25,122
1118
10,002
616
15,120
502
1,378,176
Lisbon
2,863,699
6,416,433
33,278
1993
11,571
1104
21,707
889
547,631
Helsinki
1,956,271
3,366,564
24,050
1165
3248
476
20,802
689
605,523
Dresden
1,785,632
3,803,045
7948
605
3259
340
4689
265
529,781
Tallinn
1,498,462
2,791,094
6718
499
3610
321
3108
178
424,445
Salzburg
1,272,880
2,293,208
7575
446
4735
336
2840
110
148,521
Krakow
805,664
1,846,690
9109
510
3927
317
5182
193
759,137
Bratislava
783,618
1,548,522
2859
277
1646
184
1213
93
462,603
Graz
488,681
907,964
4488
176
1131
89
3357
87
265,318
Ljubljana
425,163
794,646
3370
252
1668
167
1702
85
280,278
Correlation
Arrivals/im
ages
Arrivals/users
Bed
nights/images
Bed
nights/users
Co-efficient
0.8075
0.8375
0.8640
0.8738
Tourism Geographies 93
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
morphology of the central parts of these cities. Some of the main similarities between the
three cities can be deducted at first sight. All three have an extended historic core with
intensive tourist usage, but have also some main sites outside the centre. These isolated
attractions are either not connected at all to the centre (Vienna, Sch€onbrunn) or connectedwith just one route (Prague, Vysehrad; Budapest, City Park). The most characteristic pat-
terns in the central urban cores are axial ones related to some historical urban composi-
tions, avenues or squares; in fact these urban places are attractions themselves, and
Figure 1. Tourist (black) and local (gray) activity in Vienna (a), Prague (b) and Budapest (c)according to geotagged photography retrieved from Flickr.
94 B. K�ad�ar
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
tourists use representative historical routes instead of less appealing streets to move
between other central attractions to maximise the foreseen experience in a given time.
The historical centre of Vienna on the right bank of the Danube canal is defined by
medieval city walls that were demolished between 1858 and 1874, replaced by the urban
composition of the Ring. Photographs draw the main tourist activity inside and around
the Ring, with St. Stephen’s Cathedral marking the geometrical and tourist centre of the
city (Figure 2). A series of pedestrian streets starting here (Graben, Kerner Strasse) con-
nect to the main entry points of the Ring, the Opera and the Hofburg. The residence of
the Habsburg emperors is the most visible axial pattern ending in the recently refurbished
Museumsquartier. Vienna has important tourist sites outside the Ring like the buildings
of Hundertwasser or the Prater park; the Baroque structure of the parks of Sch€onbrunnPalace and Belvedere Palace appear as independent urban structures detached from the
main urban core (Figure 1(a)).
The medieval centre of Prague, divided into two by the Vltava river, consists of four
parts. The Castle District on the hilltop (Hrad�cany) and the Lesser Quarter (Mala Strana)
are on the western side of the river where the street patterns are much determined by the
hilly topography, drawn out clearly by the position of photographs (Figure 1(b)). On the
eastern banks the Old Town (Star�e Mĕsto) and the Jewish Quarter (Josefov) are limited
by a semi-ring road, built in 1871 to replace the original baroque town walls. Unlike in
Vienna, this ring is barely visible on the tourist images’ map, as the main places of inter-
est are organized on different historical urban axis. The Karlova is the oldest route of
medieval origin linking the Old Town from the Municipal House (in Republiky square)
to the Lesser Town and the Castle (Figure 3). Most important tourist sites are accessible
from it, culminating in the absolute tourist centre, the Old Town square. Other important
urban axis built in the nineteenth century starts from this square: the 750-meter long Wen-
ceslas square finishing at the National Museum on a small hill and the Parizsk�a Avenue
linking the sites of the Jewish quarter. The river embankments offering reflected city-
scapes also define major tourist flows; many of the institutions, which are now attractions,
were placed here in the nineteenth and twentieth centuries.
Budapest, divided into the towns of Buda and Pest by the river, Danube is in between
the layout of the other two capital cities. Buda has a Castle District very similar to the
one in Prague, and the waterfront situation of the two capitals is also similar. The medie-
val city walls of Pest surrounded a compact city of similar size than in the Austrian
capital. However, Pest followed a different development pattern when both Hungarian
and Austrian capital cities were reshaped in the late nineteenth century. The main institu-
tions in Pest are decorative landmarks scattered rather than composed together in large
compositions. The radial avenues and ring roads traced according to the master plan of
1870 all hosted one or two of these landmarks, but few are the routes connecting many
of them. Some attractions are just inserted into the immense tissue of closed urban blocks
of housing (Benko, 2011), resulting in tourist patterns that are more scattered in a larger
urban area than in Prague or Vienna. In Pest two semi-ring roads delimit the tourist city,
and similarly to Prague, the ring roads are not drawn out by tourist photography. The few
axial elements visible are the V�aci Street parallel to the Danube and ending in the main
Markethall, the Danube Promenade, connecting the Parliament and all bridges, and
Andr�assy Avenue, a representative route connecting the City Park with the centre, there-
fore reaching until the distant Heroes’ Square (Figure 4).
On the urban level the scaled map visualizations of tourist activities deducted from the
geotagged images uploaded to Flickr made the tourist patterns correlated to the historical
morphological structures of these three cities comparable.
Tourism Geographies 95
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
Correlations and comparisons at the level of the attraction
It is assumed that the direct correlation between the number of tourists visiting a city and
the number of tourist photographs taken there can be extended to the level of the attrac-
tions. The mesh of 10 by 10 meters used is refined enough to allow the analysis of tourist
Figure 2. The 30 attractions most visited by Flickr users in Vienna marked on the visualization onGoogle Earth of tourists’ geotagged photography. The images captured in 2011 are compared withofficial visitor statistics.
96 B. K�ad�ar
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
presence around specific attractions within a city. The variables associated to each Flickr
image allow a complex analysis of attendance data for singular places. First of all the
number of images or the number of users uploading these are separately comparable with
specific tourism statistics, while the analysis of the proportion of photographs and photog-
raphers can give information on the time spent at different locations. The second prospect
in the method is the possible differentiation between tourist and local users, giving an
Figure 3. The 30 attractions most visited by Flickr users in Prague marked on the visualization onGoogle Earth of tourists’ geotagged photography.
Tourism Geographies 97
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
insight into differences of urban space uses according to the time spent in a city. These
considerations allow the study and comparison of qualities like popularity, tourist–local
use and complexity of different tourist sites.
A first question is how the number of actual visitors correlates to the number of
images uploaded and to the number of users uploading images to Flickr. Sites worth cap-
turing with more photographs should offer more experiences, visual or other; therefore,
Figure 4. The 30 attractions most visited by Flickr users in Budapest marked on the visualizationon Google Earth of tourists’ geotagged photography.
98 B. K�ad�ar
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
these sites gain more from the overall space–time budget of tourists. Counting the number
of tourist photographs could be a suitable method to measure the intensity of visitor’s
activity, while counting the number of users uploading from the area of an attraction
should correlate with actual visitor attendance to singular sites. This correlation seems
easy to verify, as many institutions do measure visitor attendance. The best statistical
data are available in Vienna, where the number of visitors are registered for most tourist
attractions, published in TourMIS (2012) (Figure 2). However, this yearly data-set is
based on ticket selling, limiting its usability. Most tourists do not enter the controlled
parts of these attractions, as the most common layer of consumption in urban tourism is
the sight from outside. An example is the case of St. Stephen’s Cathedral, the main tourist
attraction of Vienna, ranking only eighth with only 560,000 tickets sold annually. At the
same time, the number of geotagged images and the number of users uploading these
verify the assumption that this should be the most visited site in Vienna. The geographical
location and symbolic importance makes this a crowded must-see place; tourists at least
pass by it, even if on a repeated visit. Even by entering inside tourists will not appear in
the statistics, as admission into the church is free; therefore, only those climbing the tow-
ers, visiting the catacomb or joining a guided tour will be recorded. The opposite happens
at the Sch€onbrunn Palace, the most visited attraction according to statistics, only the
fourth if Flickr data are considered. Tourists visit this attraction advisedly, as the palace
is situated far out from the city centre. The only activity possible nearby is to enter the
palace or visit the immense gardens, both only possible with a valid ticket, therefore all
visitors can be easily counted. Sometimes the numbers of photographs are not suitable to
measure the visits inside venues like museums as photographing is generally prohibited,
and even if allowed, geographic positioning is not a priority as most exhibitions are not
site specific. However, counting the user’s photographs seems to be a more precise tool
than official statistics to compare how many tourists did see different attractions in the
urban space. This can be further divided into groups of tourists and locals.
The five-day limit in defining tourist users divides well repeat visitors and local resi-
dents form the typical sightseeing tourists, allowing to analyse the differences of their
behaviours in space usage. Vienna has the highest ratio of local photographs compared to
tourist photographs. In fact locals’ images draw out much more morphological details
around the centre than in Prague (Figure 1). The city of Prague is smaller in size and in
population than its regional rivals, but it hosts the most tourists. As a result locally
uploaded images do not outnumber those uploaded by tourists. To balance these differen-
ces the comparison of tourist and local images/users’ quotients were used, measuring the
average photographs taken by a user at a site. Tourist users limited by their strict time
budgets have no possibilities to immerse themselves into the deeper understanding of
a complex site, or to wander into unknown parts of a city. Therefore they consume
the most obvious, taking only the images they were prepared to see (Urry, 1990).
Consequently local users take more photographs in almost every site than tourists.
The 30 tourist sites most visited by users of Flickr.com in the three cities were com-
pared according to these variables (Figures 2–Figure 4). First questions of popularity
among tourists and locals were examined, and then some observations on the visual and
functional attractiveness were made using data of the three cities.
All three cities have must-see attractions more visited by tourists than locals. The quo-
tient of tourist and local users uploading could indicate whether a site became a tourist-
only attraction, or it remained an urban space used by locals: above 1 these are touristy
places, while below 1 they are more visited by locals. In Vienna touristy places are the
main monuments inside the centre, but even around these the main public spaces tend to
Tourism Geographies 99
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
have balanced quotients around 1, and not even St. Stephen’s Cathedral reaches 2. The
most touristy attraction is the colorful building and tourist centre conceived by the artist
Hundertwasser: 1.96 times more tourists visit this than locals or repeated visitors do. In
Prague the most visited attractions have all quotients above 2. Old Town Square, the
most visited urban composition has 2.81 times more tourists than locals. Budapest is
more similar to Vienna in this prospect. Only the main attractions in the Castle District
have more than two times as many tourists than locals. More interesting are the tourist
sites where locals dominate even among the photographers. In Vienna these are the
Naschmarkt marketplace (0.83), the zoo (0.72), the Haus des Meeres aquarium (0.32),
some sites on the Danube Canal (0.65–0.71) and some historical urban squares outlying
the main axis of tourist flows. In Prague only the fort of Vysehrad is more visited by local
users (0.87), the next lowest quotients were registered at the 1980s structure of the
National Theatre (1.24) and the Wenceslas Square (1.34). In Budapest the most popular
among locals are the Margit Bridge and Island (0.77–057), Szabads�ag and Erzs�ebetSquares (0.86–0.94) and the busy Western Railway Station (0.61). It is hard to compare
these values between the three cities, as Prague has a considerably smaller population,
1.2 million compared to the 1.7 million of the other two cities, while Budapest has a lower
number of tourist visits. Still, the number of local and tourist Flickr users should correlate
with the actual number of visitors to a site, and considering this, there is a great difference
between the proportions of tourist and local users in the historical centre of Prague and the
other two cities.
The average number of photographs taken at a site gives information about some kind
of attractiveness or complexity of that site. Sites with a complex urban arrangement, with
many architectural elements and details are photographed more by an average user, while
other attractions offer only one or two good views. An extent is the case of the Viennese
Secession pavilion, an exhibition space with characteristic Art Nouveau architecture.
Only few enter, as this venue is only 30th among the most visited attractions according to
statistics, but the special decoration of the building attracts many tourists to take a look.
This is the 20th most visited attraction by Flickr users in Vienna, but the average photo-
graphs taken by a user is among the lowest, 4.71 among locals and 2.77 among tourists.
This little masterpiece is a typical one-shot attraction, like most landmark buildings. In
Prague the contemporary Dancing House of American star architect Frank O. Gehry is an
evident example. Ranking 10th in number of users visiting, only 2.42 photographs were
uploaded by an average tourist and 3.33 from an average local. In fact there are two or
three interesting angles to photograph this office building, having no other attractions
nearby. The bridges and embankments offering nice views both in Prague and Budapest
are also to be considered one-shot attractions. In fact without good zoom cameras even
the best skyline has only to offer one or two shots.
There are many attractions worth more than photographs. In the case of the main tour-
ist attraction in the Castle District of Budapest it is easy to trace why the average of pho-
tographs uploaded by users is as high as 6.88 for tourists and 8.59 for locals. This is a
multiple attraction, where the Gothic Matthias Church and the decorated nineteenth cen-
tury lookout venue of Fisherman’s Bastion are so close, that it is impossible to separate
different areas while counting the photographs taken here. In fact visitors approaching
will experience the church, the bastion, the view of the city from above and the combina-
tion of these all together. Similar reasons lie behind the high number of images in the
case of Old Town Square in Prague and St. Stephen’s Cathedral and Hofburg in Vienna.
More complicated is the case of the average of 4.68 and 9.19 photographs per user of the
nineteenth-century building of the main Synagogue in Budapest. The fact that
100 B. K�ad�ar
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
photographing in the interior of this synagogue is permitted, but an entrance ticket is
required can explain why this stand-alone monument is so much photographed by users,
and at the same time why it is only the 26th most visited attraction. The entrance fee
keeps away many visitors, but once inside, the visit offers many site-specific views of this
singular building, and the visitor wants to explore all in change of the money spent.
There are places where tourists take almost as many photographs in average than
locals. These are mostly urban situations out from the centre with an own complex urban
composition: Sch€onbrunn in Vienna (7.06 and 8.58), the Strahov Monastery (5.23 and
6.79) and the Vysehrad fort (8.32 and 7.75) in Prague or the Vajdahunyad Castle in
Budapest (4.43 and 4.73). All of these places are extensive in size, have complex urban or
architectural structures in themselves, therefore they can be captured only with multiple
images. The location outside the urban cores and the monofunctional tourist use explains
the lack of local activities and therefore the lower-than-average local photographing. Both
tourists and locals arrive to these places advisedly for the purpose of sightseeing.
The opposite phenomenon is evident in places with intense local activities combined
with a visual appeal. Here an average local user takes large number of pictures while tou-
rists take only few. Tourists do visit these places, but do not have time to immerse them-
selves into the various activities locals do. The Vienna Museumsquartier is a good
example for this. The former imperial stalls were not only refurbished to cultural func-
tions, but were also extended with two new landmark buildings in 2001. As a result, a
complex urban situation with lively public spaces was created, with many interesting
details to photograph and enjoy: baroque and modern, landmark building and temporary
street furniture. Tourists take an average of only 3.12 photographs, still this is the 12th
most visited site among these users. An average local spends many hours of leisure time
and takes an average of 17.73 photographs here, which is the second highest average after
the zoo in Sch€onbrunn. The zoo is the second most visited attraction according to official
statistics, ranking only 21st if the number of users uploading pictures is considered. Many
users of the zoo are children and families, more interested in the experience than the
visual consumption of it, therefore the number of local users cannot be correlated to
the measured visitor attendance. Those few who took photographs took many, as many of
the animals are interesting subjects, and the imperial architecture gives also many unique
viewpoints. The average number of photographs uploaded by a user is the highest: 9.18 in
the case of tourists and 28.36 for locals. Similar is the situation in the park of Margit
Island, Budapest. Karlsplatz in Vienna, Wenceslas Square in Prague, and Ferenciek
Square in Budapest also have similar proportions. These urban spaces are the main gath-
ering points for locals, where tourist functions are present but do not prevail.
Conclusions
A better tool to measure tourist activity in cities
Tourism studies are still in need of comparable data and quantitative methods to refine the
geography of urban tourism. Statistical tools can only measure controllable agents, but in
urban tourism few of these exist. We can measure visitor attendance in controlled tourist
sites such as museums, but the most interesting phenomenon and all conflicts in urban
tourism occur in the public spaces of cities. There are no good ways to count and compare
tourist flows in these spaces, where locals and tourists wander together seamlessly. New
technologies can help. GPS tracking technology can give accurate data on the spatial
movements of tourists, but their use in research so far was dependent on personal contact,
Tourism Geographies 101
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
just like the classical survey methods. The photo camera is a device essential for tourism,
and nowadays it became a tool capable of registering location-based data, which can be
shared over the internet. This possibility to manually or automatically position the photo-
graphs of places on maps opened the possibility to use the image taking obsession of tou-
rists as a tool to measure their spatial activities.
The present paper used the method of K�ad�ar and Gede (2013) to retrieve tourism-
related geographical information at city level and at the level of the singular urban attrac-
tions. The data-sets obtained from geotagged images resulted to correlate with other sta-
tistical data available to measure urban tourism in general, while at the level of singular
attractions, the number of users who uploaded images in the area of attractions apparently
correlates better to the actual number of visitors who saw these sites than the official sta-
tistics based on ticket selling. The latter measures exactly how many people could experi-
ence the complete tourist offer that the attraction has to offer, but gives not much
information on the tourist flows around these sites. Analysing the patterns of tourist activ-
ity in Vienna, Prague and Budapest resulted in some basic findings: similar morphologies
resulted in similar space usages, strong morphological barriers, composed urban struc-
tures always lead to stronger and more axial tourist patterns, while in situations without
these determining elements tourist activity spread away. The main tourist system of these
cities consists of an interconnected core system in the historical urban cores, and of singu-
lar detached elements which are visited by the average tourist only if they carry some sort
of urban complexity in themselves. Analysing and comparing the activity near the single
attractions lead to a categorization of these. There are complex sites worth capturing
numerous pictures, and one-shot attractions, like viewpoints and single monumental
buildings. There are sites with a considerable local attention measurable, and there are
the classical main attractions, where tourist users dominate. The identification and com-
parison of these sites in the urban system is much harder and inaccurate with other
methods.
Obviously this method has to be further refined. More work is needed to combine it
with other techniques of tourist tracking and to make deeper analysis of the pictures’ sub-
jects and on the users who uploaded them. Girardin et al. (2008) did research on the addi-
tional data provided by Flickr users, finding the nationalities available for many. With the
amount of data processed in this paper, such a manual analysis would be possible only
within smaller control groups. Another interesting potential in the analysis of geotagged
photography is the temporal succession of the images of users. If photographs of a user
are connected according to their time stamps on the map, these should draw out the routes
a tourist made in the urban environment. Even in this case only the data of a limited num-
ber of users could be used. Other interesting possibilities lie within the temporal analysis
of geotagged photography, many of them noted by K�ad�ar and Gede (2013). More work is
needed to study the differences in space usages at different times of the day or different
seasons of the year as well as the changes in different sites over the years. The latter could
be an important tool in the future to measure the development of urban destinations, but
today only the uploads of some 10 years are available on Flickr.
The resolution and quantity of present data already draws out the structures and inten-
sities of urban tourist space networks. Networks are comparable systems, so the possibili-
ties to make such comparisons between different cities’ tourist space systems could bring
important findings. Local and tourist land use could also be confronted more systemati-
cally, promising a better understanding of conflicts in the tourist city. The data-sets ana-
lysed suggest that most of the people using the centre of Prague are tourists, at least the
proportion between locals and tourists are much less balanced than in the other two cities.
102 B. K�ad�ar
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
This reading of Flickr photo uploads verifies the observations of many scholars, reporting
decreasing population and streets overcrowded by tourists in Prague’s centre (Cooper &
Morpeth, 1998; Deichmann, 2002; Maxwell, 1995; Simpson, 1999). Future analysis is
needed in this field to understand the connections between the number of tourists and
local users and the conflicts in the tourist-historic city, helping also the urban planning
practice to implement developments for a sustainable urban tourism.
Acknowledgements
The author is grateful for the help of M�aty�as Gede who produced the geovisualizations for thispaper, and would like to thank some anonymous reviewers and some colleagues and family mem-bers for their constructive comments on an earlier version of the manuscript.
Notes on contributors
B�alint K�ad�ar is an assistant lecturer and PhD candidate at the Department of Urban Planning andDesign of the Budapest University of Technology and Economics. His research work focuses on thedevelopment of tourism infrastructures in historic urban centres. He is involved in such develop-ments in Budapest also as an architect and urban planner.
References
Ahas, R., Aasa, A., Mark, €U., Pae, T., & Kull, A. (2007). Seasonal tourism spaces in Estonia: Casestudy with mobile positioning data. Tourism Management, 28, 898–910.
Ashworth, G. J., & Page, S. J. (2011). Urban tourism research: Recent progress and current para-doxes. Tourism Management, 32(1), 1–15.
Ashworth, G. J., & Turnbridge, J. E. (1990). The tourist-historic city. London: Belhaven Press.Benko, M. (2011). Budapest urban blocks and their sustainability. Architekt�ura & Urbanizmus,
45(3–4), 188–201.Cooper, C., & Morpeth, N. (1998). The impact of tourism on residential experience in central-
eastern Europe: The development of a new legitimation crisis in the Czech Republic. UrbanStudies, 35(12), 2253–2275.
Crandall, D., & Snavely, N. (2012). Modeling people and places with internet photo collections.Communications of the ACM, 55(6), 52.
Deichmann, J. I. (2002). International tourism and the sensitivities of central Prague’s residents.Journal of Tourism Studies, 13(2), 41–52.
Garrod, B. (2009). Understanding the relationship between tourism destination imagery and touristphotography. Journal of Travel Research, 47(3), 346–358.
Gede, M. (2012). Visualization methods of spatial distribution of geotagged photography. In E.Bircs�ak (Ed.), Data is beautiful conference (pp. 52–56). Budapest: Kitchen Budapest.
Girardin, F., Fiore, F. D., Ratti, C., & Blat, J. (2008). Leveraging explicitly disclosed location infor-mation to understand tourist dynamics: A case study. Journal of Location Based Services, 2(1),41–46.
Hayllar, B., & Griffin, T. (2005). The precinct experience: A phenomenological approach. TourismManagement, 26(4), 517–528.
Henderson, J. C. (2002). Heritage attractions and tourism development in Asia: A comparativestudy of Hong Kong and Singapore. International Journal of Tourism Research, 4, 337–344.
Hollenstein, L., & Purves, R. S. (2010). Exploring place through user-generated content: UsingFlickr to describe city cores. Journal of Spatial Information Science, 1(1), 21–48.
Jenkins, O. H. (2003). Photography and travel brochures: The circle of representation. TourismGeographies, 5(3), 305–328.
Ji, R., Gao, Y., Zhong, B., Yao, H., & Tian, Q. (2011). Mining Flickr landmarks by modeling recon-struction sparsity. ACM Transactions on Multimedia Computing, Communications, andApplications, 7S(1), 1–22.
Tourism Geographies 103
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014
K�ad�ar, B., & Gede, M. (2013). Where do tourists go? Visualizing and analyzing the spatial distribu-tion of geotagged photography. Cartographica, 48(2), 78–88.
Larsen, J. (2006). Geographies of tourist photography choreographies and performances. In J.Falkheimer & A. Jansson (Eds.), Geographies of communication: The spatial turn in mediastudies (pp. 241–258). G€oteborg: Nordicom.
Lew, A. A., & McKercher, B. (2006). Modeling tourist movements: A local destination analysis.Annals of Tourism Research, 33(2), 403–423.
Lo, I. S., McKercher, B., Lo, A., Cheung, C., & Law, R. (2011). Tourism and online photography.Tourism Management, 32(4), 725–731.
Maxwell, J. (1995). Czech and Slovak tourism patterns – problems and prospects. Tourism Manage-ment, 16(1), 21–28.
McKercher, B., & Lau, G. (2008). Movement patterns of tourists within a destination. TourismGeographies, 10(3), 355–374.
McKercher, B., Shoval, N., Ng, E., & Birenboim, A. (2012). First and repeat visitor behaviour: GPStracking and GIS analysis in Hong Kong. Tourism Geographies, 14(1), 147–161.
Modsching, M., Kramer, R., Hagen, K. T., & Gretzel, U. (2008). Using location-based tracking datato analyze the movements of city tourists. Information Technology & Tourism, 10(1),31–42.
Pang, Y., Hao, Q., Yuan, Y., Hu, T., Cai, R., & Zhang, L. (2011). Summarizing tourist destinationsby mining user-generated travelogues and photos. Computer Vision and Image Understanding,115(3), 352–363.
Pearce, D. G. (1998). Tourist districts in Paris: Structure and functions. Tourism Management,19(1), 49–65.
Pearce, D. G. (1999). Tourism in Paris, studies at the microscale. Annals of Tourism Research,26(1), 77–97.
Pearce, D. G. (2001). An integrative framework for urban tourism research. Annals of TourismResearch, 28(4), 926–946.
Pettersson, R., & Zillinger, M. (2011). Time and space in event behaviour: Tracking visitors byGPS. Tourism Geographies, 13(1), 1–20.
Popescu, A., & Grefenstette, G. (2009). Deducing trip related information from Flickr. In Proceed-ings of the 18th International Conference on World Wide Web – WWW ‘09 (pp. 1183–1184).New York, NY: ACM.
Puczk�o, L., & R�atz, T. (2000). The three forming one: Destination BPV. In J. Ruddy & S. Flanagan(Eds.), Tourism destination marketing. Gaining the competitive edge (pp. 367–377). Dublin:Tourism Research Centre, Dublin Institute of Technology.
Russo, A. P., & Van der Borg, J. (2002). Planning considerations for cultural tourism: A case studyof four European cities. Tourism Management, 23(6), 631–637.
Shoval, N. (2008). Tracking technologies and urban analysis. Cities, 25, 21–28.Shoval, N., & Isaacson, M. (2007). Tracking tourists in the digital age. Annals of Tourism Research,
34(1), 141–159.Shoval, N., & Isaacson, M. (2010). Tourist mobility and advanced tracking technologies. New
York, NY: Routledge.Shoval, N., McKercher, B., Ng, E., & Birenboim, A. (2011). Hotel location and tourist activity in
cities. Annals of Tourism Research, 38(4), 1594–1612.Shoval, N., & Raveh, A. (2004). Categorization of tourist attractions and the modeling of tourist cit-
ies: Based on the co-plot method of multivariate analysis. Tourism Management, 25(6), 741–750.
Simpson, F. (1999). Tourist impact in the historic centre of Prague: Resident and visitor perceptionsof the historic built environment. The Geographical Journal, 165(2), 173–183.
Stansfield, C. A. (1964). A note on the urban-nonurban imbalance in American recreationalresearch. Tourism Review, 19(4), 196–200.
TourMIS. (2012). City tourism in Europe. Retrieved from http://www.tourmis.infoUrry, J. (1990). The tourist gaze – Leisure and travel in contemporary societies. London: SAGE.Van der Borg, J., Costa, P., & Gotti, G. (1996). Tourism in European heritage. Annals of Tourism
Research, 23(2), 306–321.
104 B. K�ad�ar
Dow
nloa
ded
by [
Tha
mm
asat
Uni
vers
ity L
ibra
ries
] at
04:
40 0
9 O
ctob
er 2
014