Applied Geography 55 (2014) 184e192

Contents lists available at ScienceDirect

Applied Geography
journal homepage: www.elsevier.com/locate/apgeog

Evaluating the structure and use of hiking trails in recreational areas

using a mixed GPS tracking and graph theory approach

n Gonza

lez b, Xavier Garcia-Masso

c,
Karolina Taczanowska a, *, Luis-Milla

-Luis Toca-Herrera d
Andreas Muhar a, Christiane Brandenburg a, Jose
a

Institute of Landscape Development, Recreation and Conservation Planning, University of Natural Resources and Life Sciences (BOKU), Vienna, Austria
Department of Physical Education and Sport, University of Valencia, Valencia, Spain
ctica de l'Expressio
stica i Corporal, University of Valencia, Spain

Musical, Pla
Departament de Dida
d
Institute for Biophysics, University of Natural Resources and Life Sciences (BOKU), Vienna, Austria
b
c

a r t i c l e i n f o

a b s t r a c t

Article history:
Available online

Recreational trails encourage numerous outdoor leisure activities in a variety of urban, rural, and natural
environments. Understanding the way trails function is crucial for the designers and managers of recreational sites to balance the needs of visitors and site capacities. This paper presents a new approach to
evaluate the structure and use of hiking trails by combining GPS tracking and analytical methods based
on graph theory. The study is based upon empirical data (N 482 GPS tracks) collected in the Lobau,
which is part of the Danube Floodplains National Park in Austria. The physical structure of trails
(structural network; undirected graph) and their usage (functional network; directed graph) were
analysed using a graph theory approach. The network coherence (connectivity indices: b, g, a), the
movement direction at path segments and the relative importance of network nodes (node centrality
measures: degree, closeness, betweenness) were calculated. The Lobau trail network is not evenly used
by park visitors. Therefore, the calculated parameters differ between the structural and functional networks. From management perspective the results obtained for the functional network are particularly
important. 61% of recreational use (hiking) concentrates on designated trails, 21% on non-marked paths
and 18% is off-trail use. In most cases the location of signposts and information boards in the Lobau
corresponds with the highest node centrality measures in the functional network (degree and
betweenness). The proposed methodology can be easily adopted for the evaluation of any trail network
in outdoor recreational sites.
2014 Published by Elsevier Ltd.

Keywords:
Trail management
Recreational trails
Graph theory
GPS tracking
Outdoor recreation

Introduction
Walking and hiking for well-being and tness are major outdoor
recreational activities in western societies (Eurostat, 2012; Statista,
2014). Thirty minutes of walking a day is considered to have a
positive impact on human health and well-being (WHO, 2010).
Increasingly, the relationships between the physical environment
and physical activity are being identied (Brownson, Hoehner, Day,
Forsyth, & Sallis, 2009). Numerous studies show that accessibility
and the quality of recreational areas encourage physical activity
(Handy, Boarnet, Ewing, & Killingsworth, 2002; Owen, Humpel,

* Corresponding author.
E-mail addresses: karolina.taczanowska@boku.ac.at,
(K. Taczanowska).
http://dx.doi.org/10.1016/j.apgeog.2014.09.011
0143-6228/ 2014 Published by Elsevier Ltd.

karolt22@yahoo.com

Leslie, Bauman, & Sallis, 2004; Saelens & Handy, 2008). In

densely populated areas such as large cities, the provision of high
quality outdoor leisure areas is a particularly important issue.
Balancing the needs of visitors and site capacities is a major challenge for the planners and managers of such amenities (Bell, 1997).
This is particularly important in areas of high natural value that are
threatened by negative impacts from unbalanced use. A comprehensive understanding of human spatial behaviour and the actual
use of recreational infrastructure can support the effective planning
of outdoor leisure settings (Cessford & Muhar, 2004). The strategic
allocation of infrastructure and provision of information are
important tools to manage the direction of movement and the
recreational experience of visitors. There is a respective body of
research on recreational trail design, trail use and trail ecology
(Arrowsmith & Inbakaran, 2002; Beeco, Hallo, English, & Giumetti,
2013; Dixon, Hawes, & McPherson, 2004; Hawes, Dixon, & Ling,

K. Taczanowska et al. / Applied Geography 55 (2014) 184e192

185

Fig. 1. Location of the case study area e the Lobau.

2013; Tomczyk & Ewertowski, 2013) however, most of the studies

focus on path attributes rather than on the topological relationships
between network components.
Graph theory has so far been used in this eld to assess the
overall connectivity in existing trail networks (Kolodziejczyk,
2011; Li, Ge, & Liu, 2005; Styperek, 2001). It is believed that
the high connectivity of paths contributes to an efcient ow of
individuals between different locations (Gross & Yellen, 2006)
and may therefore enhance the recreational opportunities for
visitors.
The aim of this paper is to present a new method of evaluating
the functionality of a recreational trail network by combining GPS
tracking and graph theory. The main focus is on the assessment of
the overall trail connectivity as well as the relative importance of
network nodes. The study compares the topological properties of a
structural and functional network in the case study area. The term
structural network refers to the physical layout of all paths in the
study area, whereas the term functional network refers to the
actual use of the paths by recreationists.
In this way, the study aims to complement and enhance a
traditionally used GIS-based network analysis (Hinterberger,

Arnberger, & Muhar, 2002; Taczanowska, 2009), where only the

volumes of visits per path segment without considering any topological relationships between network components are being
considered.
Materials and methods
Case study area
The study is based on empirical data collected in the Lobau
recreational area in Austria. The area is part of the Danube Floodplains National Park and is located within the borders of the capital
city of Austria, Vienna (Figs. 1 and 2). This recreational area
(2400 ha, 2 km wide  10 km long) attracts approximately 600,000
visits per year (Brandenburg, 2001). Hiking and cycling are the two
major recreational activities practiced in the area (Brandenburg,
2001). The forecasts for the next decades anticipate a signicant
increase in visitation levels due to residential developments in the
vicinity (Arnberger et al., 2012), which will cause increased pressure on the natural environment and pose a challenge for the National Park management.

Fig. 2. Network of recreational trails and paths in the Lobau.

186

K. Taczanowska et al. / Applied Geography 55 (2014) 184e192

Data collection
Our study is based upon two types of data: the physical structure of a trail network and a record of trip itineraries (GPS tracks) of
recreational area visitors. The trail network has been digitised from
existing trail maps into the ESRI shapele format (Hinterberger,
2000), differentiated by trail characteristics (marked trails, nonmarked paths) and veried using GPS during eldwork.
To investigate the actual use of trails in the Lobau, a GPS tracking
approach was applied (Taczanowska, Muhar and Brandenburg
2008). In total, 482 trip itineraries (GPS tracks) of individual visitors were collected during 4 sampling days in the case study area.
The selected sampling days depicted a typical good weather
weekend situation in the most heavily used seasons (spring and
autumn). During the four sampling days, a total number of 5472
hikers entered the Lobau area through four major entrances. In
total, 60 GPS devices (GARMIN e-Trex) were distributed each day at
the entrance gates to randomly selected hikers. Visitors performing
other activities such as cycling or jogging were not included in this
study. The route of each study participant was recorded as a set of
track points with a constant temporal interval of 4 s.
Data analysis
Data analysis consisted of ve steps: 1) Pre-processing of GPS
data; 2) Creating the structural network (undirected graph); 3)
Creating the functional network (directed graph); 4) Calculating the
network connectivity indices; and 5) Quantifying the importance of
network nodes (node centrality measures). The pre-processing of
the collected GPS data and the analysis of the structural and
functional trail networks were conducted using a Matlab modelling
environment. The parameters describing the importance of nodes
and the graphic representation were calculated using the Pajek
software system (Batagelj & Mrvar, 2003).
Pre-processing of GPS data
In a rst step, anomalous GPS data due to bad signal quality
and unrealistic movement distances or speeds were detected and
ltered. Limited GPS signal reception can occur under dense forest

cover; we eliminated GPS tracks where the signal interruption

was longer than 1 min (>15 trackpoints) or where the total sum of
track sections with poor data quality was more than 500 m. We
also eliminated records with movement speeds outside the usual
range of hiking. The normal walking speed for healthy humans
(between 20 and 99 years) oscillates between 2 km/h and 6 km/h
(Bohannon & Williams Andrews, 2011). If the visitor runs, the
speed range can vary from 11.5 km/h to 15.2 km/h (Cavanagh &
Kram, 1989). Finally, only 314 GPS tracks ( 65% of the records)
were taken into account for the construction of the functional
network and further analysis.
Theory and calculation
Graph theory is applicable to many questions related to the
properties and function of networks (Gross & Yellen, 2006). It has
been successfully used in many application domains such as
transportation (Rodrigue, Comtois, & Slack, 2013; Taaffe &
Gauthier, 1973), landscape ecology (Urban & Keitt, 2001) and sociology (Wasserman & Faust, 1994). A graph (G) is described as a
nonempty set of vertices (V) and a set of edges (E), where each edge
determines a set of two vertices called the endpoints of the edge
(Richmond & Richmond, 2004):

G VG; EG
A directed graph is a nonempty set of vertices and a set of
directed edges, where each directed edge determines an ordered
pair of vertices (initial vertex and terminal vertex).

G VG; EG
where u; v2VG  VG
In this paper the terms vertex and node are being used
synonymously. The same applies to the terms edges and links
which have here the same meaning.
Creating the structural network
The GIS-based inventory of the Lobau trail network was imported into the Matlab environment. The undirected graph of the

Fig. 3. Example of a directed graph based on collected trip itineraries (GPS tracks) in the Lobau area. (A) An example of a recorded route of a hiker in the Lobau area. Red circles
indicate GPS track points; blue points indicate the network nodes (vertices) where the visitor has a choice of different walking paths. The blue squares refer to the maximum
accepted range to go through the node. (B) Representation of the corresponding graph; G (V,E) where the vertices are V {1,,8} and the edges E {(1, 2), (2, 3), (3, 4), (4,5), (5,6),
(6,7), (7,7), (7,8), (8,3), (3,2), (2,1)}. In the nal graph, the loop (7,7) was deleted because the property of directed graphs is vi s vj. (For interpretation of the references to colour in
this gure legend, the reader is referred to the web version of this article.)

K. Taczanowska et al. / Applied Geography 55 (2014) 184e192

187

Table 1
Centrality measures used to characterise the relative importance of nodes within the undirected and directed graphs.
Centrality

Description

Input degree

Number of edges entering to vertex i.

Mathematical equation
P
d

xij
i

Output degree

Number of edges leaving vertex i.

d
i

j2G

xij

j2G

Degree (all)
Closeness

Total number of edges connected to the vertex i.

Inverse sum of distances from a given vertex to all other vertices in the graph

Di d d
P
Ci
di; j1

Betweenness

A number of times a vertex is crossed by shortest paths in the graph

Bi

jsi2G

j<k

gjk i=gjk

where gjk the number of geodesics connecting jk,

and gjk(i) the number that geodesics i is on.
Mathematical equations based on (Rodrigue et al., 2013).

structural network was stored as G (V(G),E(G)), where

V {v1,,vn} is the set of vertices (nodes) of the trail network, and
E {e1,,em} is the set of edges (physically existing paths between
vertices), considering that every ek belongs to {vi, vj} with vi, vj that
belongs to V, vi s vj.

were eliminated from the functional network. Table 1 provides an

overview of the calculated centrality measures.
Results
General characteristics of the structural and the functional network

Creating the functional network (directed graph)

The directed graph was built following a procedure similar to
the undirected graph (see Section Creating the structural network),
but the coordinate pairs were stored as edges. Here, the rst
number represented the node from where the visitor left, whereas
the second number was related to where the visitor arrived.
The functional network was based on the collected GPS tracks.
Fig. 3 shows an example of a directed graph generated for a single
route completed by a visitor. The functional network is the result of
all the visitors' walking paths between the nodes of the trail network.
Calculating network connectivity indices
To evaluate the coherence of the structural and functional networks, the following connectivity measures were calculated. For
more detailed description see (Rodrigue et al., 2013):
- Kansky index b e=v, where e is the number of edges, and v is
the number of vertices; the higher the value of b, the more
coherent the network.
- Kansky index g e=3*v  2, dening the ratio of the existing
number of edges (e) to the maximum possible number of edges
resulting from a number of vertices (v). The value of g is between 0 and 1, where the value of 1 indicates a completely
connected network and is very unlikely in reality.
- Kansky index a m=2v  5, wherem is a cyclomatic measure
calculated as m e  v p, where p is the number of isolated
subgraphs. This index describes the meshedness of a graph; 1
indicates a completely meshed network, whereas 0 indicates a
very simple network.

Calculating node centrality measures

To quantify the relative importance of nodes within the Lobau
trail network, ve different centrality measures were calculated:
input degree, output degree, degree all, closeness and betweenness. The centrality measures were calculated for both the structural and functional networks. In the functional network, nodes
without activity were not taken into account. In addition, loops

The Lobau trail network consisted of 268 vertices (nodes) and

405 edges. Fig. 4 illustrates the obtained graphs representing the
structural and functional networks in the Lobau area.
The investigated trail network was not evenly used by the
hikers. The visitors walked through 222 nodes using 155 paths
(edges). The recreational use is concentrated in the northern and
western part of the area, whereas the southeastern part of the
Lobau (Untere-Lobau) is less frequently used. Visitors tended to use
trails located close to the major entrance points. In total, 23.2% of
the visitors entered the park though node number 1 (Dechantweg;
located in the western part of the Lobau near a visitor centre and a
parking lot). The other important nodes used to access the park
were numbers 47 (22.5%), 86 (16.8%) and 161 (11.1%).
The analysis shows that only 61% of the used network edges
(functional network) had a designated trail status in the structural
network. Approximately one fth of the used path segments were
not marked, and 18% of the functional network connections indicated off-trail use (movement of hikers outside the structural trail
network). Table 2 provides an overview of the used path segments
grouped by the path type.
In addition, the functional network contains information about
walking directions in the investigated area. The walking directions
observed at each path segment are presented in Fig. 4. Most of the
paths were equally used in both directions; however, there were
also trail segments with one dominant direction of use. For
example, the interpretative trail Naturlehrpfad Obere Lobau was
used in an anti-clockwise direction.
Network connectivity
In general, we observed that the network of the existing paths in
the Lobau had moderate connectivity (b 1.51; g 0.50; a 0.25).
It belongs to the lattice network type (Taaffe & Gauthier, 1973),
which is characterised by a moderate density of edges relative to
the number of nodes. The network of designated trails in the Lobau
was less coherent (b 0.61; g 0.21). The connectivity indices of
the actually used trails (functional network) in the Lobau were
lower than those in the entire structural path network (b 0.70;
g 0.24; a 0.14). However, the calculated indices were slightly
higher than the results obtained for the marked path network. This
conrms previous results that showed that only a part of the

188

K. Taczanowska et al. / Applied Geography 55 (2014) 184e192

Fig. 4. Structural trail network and functional network. Part A shows the structure of the Lobau trail network, composed of 268 nodes and 405 edges. The functional network,
created from the paths used by the visitors, is presented in part B. Dark thick arrows indicate the most visited paths. Thin and bright arrows refer to the paths that were visited less
frequently. Nodes and paths not used by visitors do not appear in the map.

existing paths were used by the hikers and that the visitors did not
exclusively stay on the marked trails.
Node centrality measures
The calculated node centrality measures showed clear differences in the importance of nodes within the investigated trail
network. The results are very heterogeneous and reect different
approaches to measure node centrality. Each of the investigated
parameters generated a slightly different distribution of node
importance within the Lobau network. Fig. 5 presents maps of the
calculated node centrality measures for the structural and functional path networks in the Lobau. Table S1 and Table S2 included in

the supplementary information of the paper provide an overview of

all centrality values obtained for the nodes of the structural and
functional networks.
The degree values obtained for nodes of the structural network
range from 1 to 5 and depict the number of paths that met at a node
location. The measure of degree in the functional network was
more heterogeneous and reected the intensity (degree) and the
direction of use (input and output degree) at each node location. As
there was a clear disproportion of use between the NW and the SE
of the park, the highest degree values were observed at the heavily
used network nodes located in the NW. The directions of use were
balanced at most node locations, which indicates that they have
mostly a transit function, where the same number of visitors enter

K. Taczanowska et al. / Applied Geography 55 (2014) 184e192

Table 2
Summary of path usage grouped by different path types.
Path type

Frequency of used path segments (sum)

Marked trail
Non-marked path
Off-trail

3148
1086
924

%
61
21
18

Total

5158

100

189

and leave a specic node. According to the collected data, there

were no nodes that were exclusively attractors or senders in the
area.
Closeness centrality is based on the network distance between
nodes (number of nodes between a specic node and other nodes).
In the structural network, the mean value of network distances
between nodes was 10.97; the largest distance was observed between node 14 and node 268, which were separated by 27
nodes. The nodes with the highest closeness values were located in
the southern-central part of the Lobau.

Fig. 5. Distribution of the node centrality measures in the structural and functional path networks in the Lobau (Degree, Closeness and Betweenness).

190

K. Taczanowska et al. / Applied Geography 55 (2014) 184e192

In the functional network, the distribution of closeness centrality was slightly different e the most important nodes were
located in the northern-central part of the park, close to the
entrance point Uferhaus. The largest distance observed in the
functional network was the distance between node 1 and node 239,
which were separated by 36 nodes.
Betweenness centrality provides information about the number
of communication paths crossing a specic node. In the structural
network, the highest betweenness values were observed mostly in
the southern-central part of the Lobau as well as near the entrance
Saltenstrasse. In the functional network, the nodes located in the
northern-central and northern part played a more important role
(entrances Uferhaus and Esslinger Furt).
Discussion
Meaning of the ndings
The main contribution of this study is the combination of GPS
tracking and analytical methods based on graph theory to better
understand the structure and function of a trail network in an
outdoor recreational site.
Typically, GIS-based network analysis is being used to study the
usage levels (volume of visits per path segment within a specied
time period) in recreational networks (Hinterberger et al., 2002;
Taczanowska, 2009). Recently, also GPS tracking has become a popular method to gather information about trip itineraries of recreationists (Beeco & Brown, 2013; Orellana & Wachowicz, 2011; Shoval
& Isaacson, 2007; Taczanowska et al., 2008) and frequently density
analysis of GPS trackpoints are being used to depict the intensity of
recreational use (Beeco & Brown, 2013; Taczanowska et al., 2008).
Graph theory, despite its established position in other elds, is
only marginally used within the outdoor recreation context. Most
of the studies using a graph-theoretic approach concentrate only on
the physical structure of trails and disregard their function
(Kolodziejczyk, 2011; Li et al., 2005; Styperek, 2001).
Our study combines both approaches: graph theory and GPS
tracking and proposes new way to study visitor ows and evaluate
recreational trail networks.
Comparison of the structural and the functional trail network in the
study area
We observed that the network of the existing paths in the Lobau
was not evenly used by visitors hiking in the area, and therefore the
parameters characterising the structural and functional networks
differ from each other. The network connectivity indices calculated
for all existing paths in the Lobau place the area in the category of
lattice networks. However, the network of the designated trails is
less coherent. The network connectivity indices of the functional
network were between the results obtained for all the paths in the
Lobau and those obtained for the marked trails only. This indicates
that the marked trail network does not fully correspond with the
behaviour of hikers, and a dense network of unofcial paths encourages hikers to step off the designated trails.
The node centrality values obtained for the structural network
and the functional network differed from each other. The degree
index directly reects the number of visitors using a particular
node. As the use of the trail network was not distributed homogenously, the differences between the structural and functional
network are substantial in this case. Large differences in the relative
importance of network nodes were observed for the betweenness
centrality values. The calculated betweenness of the structural
network indicates the importance of the nodes located in the south
and in the centre of the trail network, whereas the results obtained

for the functional network expose the importance of northern

nodes. The closeness measure is positively correlated with the
betweenness measure.
Comparison to other recreational areas
According to the authors' knowledge and the available literature, there are a limited number of studies that use graph theory to
analyse recreational trail networks. Previous studies have typically
concentrated on the connectivity measures of structural networks
and related them to different management questions such as the
optimisation of trail design (Kolodziejczyk, 2011) or tourism impact
assessments (Li et al., 2005).
The connectivity indices of the entire Lobau path network
(marked and non-marked paths) are comparable with the results
obtained for other outdoor recreational areas (Kolodziejczyk, 2011;
Li et al., 2005; Styperek, 2001). For example, the reported values of
the Kansky index b range from 1.32 (Tatra National Park, Poland) up
to 1.59 (Chelmy Landscape Park, Poland). The Kansky index g values
range from 0.33 (Multi-Colorful Pond area e Jiuzhaigou Biosphere
Reserve, China) up to 0.50 (Panda Lake, Jiuzhaigou Biosphere
Reserve, China). However, the designated trail network of the Lobau
was less coherent than the abovementioned locations.
Studies investigating the relative importance of functional
network nodes are less frequent in this application domain. Node
centralities are being used to evaluate and analyse larger structures,
such as tourism destinations (D'Agata, Gozzo & Tomaselli 2013). In
this respect, the methodology proposed in this paper is new in the
eld of outdoor recreation research.
Limitations of the proposed methodology
The main strength of this work relies on the use of GPS data to
analyse not only the static nature of the path network, but also the
way it is being used. In this way, the graph theoretic measures and
indices obtained for the entire physically existing path network can
be compared with the calculations performed for the actually used
trails.
Nevertheless, there is still room to improve the methodology.
The rst limitation refers to the use of GPS tracking to collect data
on visitor distribution within trail networks. Although GPS devices
have been widely utilised to capture trip itineraries in recreational
areas, it is common that certain problems arise during the data
collection process (e.g. partial signal loss due to heavy tree cover).
For our study we also have to consider that there might be some
spatial bias in the data due to different degrees of forest cover in the
park, as tracks passing through dense forests were more likely to be
removed from the sample due to bad GPS signal reception. Several
GPS-based studies have addressed this matter (Le Faucheur et al.,
2007; Meijles, de Bakker, Groote, Barske, 2014; Taczanowska
et al., 2008). There are approaches that aim to recover missing
data using interpolation algorithms (Wiehe et al., 2008) or other
data recovery methods. However, this aspect is beyond the scope of
the present study.
The second critical point refers to the small number of sampling
days when GPS data were collected. As visitor ows in the Lobau
change over time (daily, weekly and seasonal changes) and strongly
depend on weather conditions (Brandenburg, 2001), there is a need
to collect larger samples of GPS tracks to depict the differences in
trail use. At present, the functional trail network represents an
average Sunday situation with pleasant weather conditions. In
addition, an extrapolation of GPS data to the total number of visitors observed in the area could be considered for future work.
Furthermore, only basic graph theory parameters, such as connectivity indices (b, g, a) and node centrality measures (degree,

K. Taczanowska et al. / Applied Geography 55 (2014) 184e192

closeness, betweenness), were calculated in this study. There is a

need to explore further possibilities to measure network connectivity and node centrality and to assess their utility for recreational
area management. In addition, to advance the calculations, metric
distances between nodes could be considered by using spatial
graphs (Fall, Fortin, Manseau, & OBrien, 2007).

191

with the highest centrality values were mostly well equipped with
signposts and information boards. This means, the National Park
management has to search for alternative measures to encourage
use of the designated trails.
The proposed methodology may be easily adopted for the
evaluation of recreational trail networks in any outdoor leisure
setting worldwide.

Implications for planning and management of recreational sites

Acknowledgements
Our results can be very valuable for the planning and management of outdoor recreational areas. Connectivity indices provide
information about the actual coherence of a trail network and
enable comparisons with other areas. Connectivity indices may also
be used to quantify the evolution of a network over time (Rodrigue
et al., 2013). Previous studies have reported strong negative correlation between the connectivity of trails and trampling problems
(Li et al., 2005). The higher the trail connectivity, the less trampling
problems are observed in the area. However, the reported connectivity indices for recreational networks usually place them in
the category of lattice networks, which are characterised by a
moderate coherence. Delta network systems (very high connectivity) are more common in urban environments (Taaffe &
Gauthier, 1973). In the eld of urban planning, connectivity measurements are often correlated with the number of people using a
street network (Jiang, 2009). This relationship may be used for
trafc modelling purposes. In our case, this aspect could be
extended in future work.
Assessing the relative importance of network nodes is particularly important from a trail management perspective. Information
about key nodes in the trail network can support the design and
allocation of sign posts, on-site maps, interpretative trails and other
infrastructure. Also visitor counting devices in the course of a
monitoring program would typically be installed on such nodes.
Node centralities calculated for the functional network simultaneously take into account visitor use and the topological relationships between network components. Therefore, this indication of
the most important nodes in the trail network can be particularly
valuable. Information about the directions of visitor ows at specic path segments can also support decisions concerning infrastructure as well as social conict management. It has been shown
that the direction of movement has a signicant inuence on
crowding perception (Reichhart & Arnberger, 2010).
Conclusions
We summarise that the combination of GPS tracking and a graph
theory analytical approach can deliver important information
related to the structure and use of outdoor recreational areas.
The proposed methodology complements the traditionally used
GIS-based network analysis. Our work shows how topological
network properties may be used for trail network evaluation. We
highlight the benets of analysing not only the physical structure,
but also the function of recreational trail networks.
The calculated parameters can be practically used for trail
management in the study area. The results show that the investigated trail network does not facilitate the effective movement of
visitors between the two major parts of the area: Obere Lobau and
Untere Lobau, which is positive from a nature conservation point of
view. The NW-part of the case study area e Obere Lobau is heavily
used and a limited connectivity in the central part of the area limits
access to the SE-part of the Lobau (Untere Lobau).
Visitors' spatial behaviour did not fully correspond with the
network of designated trails in the Lobau. Only 61% of recreational
use (hiking) concentrated on designated trail network and hikers
tended to use non-marked paths or even stepped off-trail. Nodes

The GPS data were collected in the context of the BALANCE

research project within the framework of the Austrian Space Programme (ASAP). We wish to acknowledge and thank Ramona van
Marvijk and Rene Jochem (Wageningen University, The
Netherlands) for providing the GPS devices used in this study. Data
analysis was supported by a research grant of the University of

n; UniValencia, Spain (Ayudas para estancias de investigacio
versidad de Valencia: UV-INV EPDI13-113269).
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.apgeog.2014.09.011.
References
Arnberger, A., Deussner, R., Eder, R., Hein, T., Illedits, A., Kemptner, I., et al. (2012).
Perspective LOBAU 2020. Endbericht. Exploring management options of a heavily
used urban biosphere reserve confronted with new urban developments in its
neighbourhood considering a restricted potential for ecosystem development 150.
Vienna.
Arrowsmith, C., & Inbakaran, R. (2002). Estimating environmental resiliency for the
Grampians National Park, Victoria, Australia: a quantitative approach. Tourism
Management, 23, 295e309.
Batagelj, V., & Mrvar, A. (2003). Pajek e analysis and visualization of large networks.
In M. Jnger, & P. Mutzel (Eds.), Graph drawing software (pp. 77e103). Berlin:
Springer.
Beeco, J. A., & Brown, G. (2013). Integrating space, spatial tools, and spatial analysis
into the human dimensions of parks and outdoor recreation. Applied Geography,
38, 76e85.
Beeco, J. A., Hallo, J. C., English, W. R., & Giumetti, G. W. (2013). The importance of
spatial nested data in understanding the relationship between visitor use and
landscape impacts. Applied Geography, 45, 147e157.
Bell, S. (1997). Design for outdoor recreation. London: Spon.
Bohannon, R. W., & Williams Andrews, A. (2011). Normal walking speed: a
descriptive meta-analysis. Physiotherapy, 97, 182e189.
Brandenburg, C. (2001). Erfassung und Modellierung von Besuchsfrequenzen in
Erholungs- und Schutzgebieten e Anwendungsbeispiel Nationalpark DonauAuen, Teilgebiet Lobau. Institut fr Freiraumgestaltung und Landschaftspege.
t fr Bodenkultur (PhD thesis).
Wien: Universita
Brownson, R. C., Hoehner, C. M., Day, K., Forsyth, A., & Sallis, J. F. (2009). Measuring
the built environment for physical activity: state of the science. American
Journal of Preventive Medicine, 36. S99e123 e12.
Cavanagh, P. R., & Kram, R. (1989). Stride length in distance running: velocity, body
dimensions, and added mass effects. Medicine and Science in Sports and Exercise,
21, 467e479.
Cessford, G., & Muhar, A. (2004). Monitoring options for visitor numbers in national
parks and natural areas. Journal for Nature Conservation, 11, 240e250.
D'Agata, R., Gozzo, S., & Tomaselli, V. (2013). Network analysis approach to map
tourism mobility. Quality & Quantity, 47, 3167e3184.
Dixon, G., Hawes, M., & McPherson, G. (2004). Monitoring and modelling walking
track impacts in the Tasmanian wilderness world heritage area, Australia.
Journal of Environmental Management, 71, 303e318.
Eurostat. (2012). Time spent, participation time and participation rate in the main
activity by sex and age group e collection round 2000.
Fall, A., Fortin, M.-J., Manseau, M., & O'Brien, D. (2007). Spatial graphs: principles
and applications for habitat connectivity. Ecosystems, 10, 448e461.
Gross, J. L., & Yellen, J. (2006). Graph theory and its applications. Boca Raton:
Chapman & Hall/CRC.
Handy, S. L., Boarnet, M. G., Ewing, R., & Killingsworth, R. E. (2002). How the built
environment affects physical activity: views from urban planning. American
Journal of Preventive Medicine, 23, 64e73.
Hawes, M., Dixon, G., & Ling, R. (2013). A GIS-based methodology for predicting
walking track stability. Journal of Environmental Management, 115, 295e299.
Hinterberger, B. (2000). Besucherstromanalyse im Wiener Anteil des Nationalpark
Donau-Auen, der Lobau: Routenanalyse mit GIS. Institut fr Freiraumgestaltung
und Landschaftspege. Wien: Universit
at fr Bodenkultur in Wien (MSc thesis).

192

K. Taczanowska et al. / Applied Geography 55 (2014) 184e192

Hinterberger, B., Arnberger, A., & Muhar, A. (2002). GIS-supported network analysis
of visitor ows in recreational areas. In A. Arnbrger, C. Branderburg, & A. Muhar
(Eds.), Monitoring and management of visitor ows in recreational and protected
areas (pp. 28e32). Vienna, Austria: Institute for Landscape Architecture and
Landscape Management, BOKU University.
Jiang, B. (2009). Ranking spaces for predicting human movement in an urban
environment. International Journal of Geographical Information Science, 23,
823e837.
Kolodziejczyk, K. (2011). Hiking trails for tourists in the Chelmy landscape park e
assessment of their route and infrastructure development. Polish Journal of
Sport and Tourism, 18, 324e329.
Le Faucheur, A., Abraham, P., Jaquinandi, V., Bouye, P., Saumet, J. L., & NouryDesvaux, B. (2007). Study of human outdoor walking with a low-cost GPS and
simple spreadsheet analysis. Medicine and Science in Sports and Exercise, 39,
1570e1578.
Li, W., Ge, X., & Liu, C. (2005). Hiking trails and tourism impact assessment in
protected area: Jiuzhaigou biosphere reserve, China. Environmental Monitoring
and Assessment, 108, 279e293.
Meijles, E. W., de Bakker, M., Groote, P. D., & Barske, R. (2014). Analysing hiker
movement patterns using GPS data: implications for park management. Computers, Environment and Urban Systems, 47, 44e57.
Orellana, D., & Wachowicz, M. (2011). Exploring patterns of movement suspension
in pedestrian mobility. Geographical Analysis, 43, 241e260.
Owen, N., Humpel, N., Leslie, E., Bauman, A., & Sallis, J. F. (2004). Understanding
environmental inuences on walking; review and research agenda. American
Journal of Preventive Medicine, 27, 67e76.
Reichhart, T., & Arnberger, A. (2010). Exploring the inuence of speed, social,
managerial and physical factors on shared trail preferences using a 3D computer animated choice experiment. Landscape Urban Plan, 96, 1e11.
Richmond, B., & Richmond, T. (2004). A discrete transition to advanced mathematics.
Belmont, CA: Thomson/Brooks/Cole.
Rodrigue, J.-P., Comtois, C., & Slack, B. (2013). The geography of transport systems.
New York: Routledge.

Saelens, B. E., & Handy, S. L. (2008). Built environment correlates of walking: a

review. Medicine and Science in Sports and Exercise, 40, S550eS566.
Shoval, N., & Isaacson, M. (2007). Tracking tourists in the digital age. Annals of
Tourism Research, 34, 141e159.
Statista. (2014). The statistics portal. Number of participants in hiking in the U.S. from
2006 to 2012.
Styperek, J. (2001). Tourist hiking trails in Polish National Parks [in Polish]. Turyzm,
11, 25e37.
Taaffe, E. J., & Gauthier, H. L. (1973). Geography of transportation. Englewood Cliffs,
N.J.: Prentice-Hall.
Taczanowska, K. (2009). Modelling the spatial distribution of visitors in recreational
areas, 199. Vienna, Austria: University of Natural Resources and Life Sciences
(BOKU) (PhD thesis).
Taczanowska, K., Muhar, A., & Brandenburg, C. (2008). Potential and limitations of
GPS tracking for monitoring spatial and temporal aspects of visitor behaviour in
recreational areas. In A. Raschi, & S. Trampetti (Eds.), Proceedings of the Fourth
International Conference on Monitoring and Management of Visitor Flows in
Recreational and Protected Areas. MontecatiniTerme 14e19 Oct. 2008 (pp.
451e456).
Tomczyk, A. M., & Ewertowski, M. (2013). Planning of recreational trails in protected
areas: application of regression tree analysis and geographic information systems. Applied Geography, 40.
Urban, D., & Keitt, T. H. (2001). Landscape connectivity: a graph-theoretic
perspective. Ecology, 82, 1205e1218.
Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications.
Cambridge ; New York: Cambridge University Press.
WHO. (2010). Global recommendations on physical activity for health. Geneve: World
Health Organisation.
Wiehe, S. E., Hoch, S. C., Liu, G. C., Carroll, A. E., Wilson, J. S., & Fortenberry, J. D.
(2008). Adolescent travel patterns: pilot data indicating distance from home
varies by time of day and day of week. The Journal of Adolescent Health: Ofcial
Publication of the Society for Adolescent Medicine, 42, 418e420.