Biodiversity, scenery and infrastructure: Factors driving wildlife tourism in an African savannah national park

Biodiversity, scenery and infrastructure: Factors driving wildlife tourism in an African savannah national park

Biological Conservation 201 (2016) 60–68 Contents lists available at ScienceDirect Biological Conservation journal homepage: www.elsevier.com/locate...

667KB Sizes 132 Downloads 258 Views

Biological Conservation 201 (2016) 60–68

Contents lists available at ScienceDirect

Biological Conservation journal homepage: www.elsevier.com/locate/bioc

Biodiversity, scenery and infrastructure: Factors driving wildlife tourism in an African savannah national park Claudia Grünewald a,b,⁎, Matthias Schleuning a, Katrin Böhning-Gaese a,b a b

Senckenberg Biodiversity and Climate Research Centre (BiK-F), Senckenberganlage 25, 60325 Frankfurt, Germany Institute for Ecology, Evolution and Diversity, Department of Biological Sciences, Goethe University, Max-von-Laue Straße 13, 60438 Frankfurt, Germany

a r t i c l e

i n f o

Article history: Received 21 December 2015 Received in revised form 20 May 2016 Accepted 31 May 2016 Available online xxxx Keywords: Ecosystem service Visitor numbers Visitor interviews Large mammals Visibility Landscape

a b s t r a c t Wildlife tourism is an important cultural ecosystem service, benefiting regional economies and biodiversity conservation. Many wildlife tourism destinations remain below their visitor and income capacities. Management strategies are needed that increase visitor satisfaction and a destination's reputation to attract more visitors. Wildlife tourism can be directly linked to biodiversity, but might also be directly and indirectly influenced by other factors, such as landscape features or infrastructure. We investigated the relationships between visitor numbers and biodiversity, along with other factors, in a major wildlife tourism destination using structural equation modeling and additionally assessed visitors' expectations and viewing preferences. We simultaneously recorded large mammal and visitor data along 78 road transects in Kruger National Park (KNP), South Africa, and conducted interviews with visitors. We also collected data on vegetation cover, visibility, landscape features and infrastructure. We found high visitor numbers at transects with high sighting probabilities of large predators, while other factors, e.g. ungulate densities or infrastructure, were only weakly associated with visitor numbers. Consistently, interview results suggested that seeing wildlife was the main reason for visiting the park, and large predators, especially lions and leopards, ranked highest among the visitors' wildlife preferences. Our results demonstrate that wildlife tourists in KNP are primarily attracted to large predators. To meet visitor expectations and to increase visitor numbers, park management should focus on the conservation of natural savannah ecosystems with large predator and prey populations. With such an ecosystem-based management, biodiversity conservation can be successful while securing wildlife tourism and its revenues. © 2016 Elsevier Ltd All rights reserved.

1. Introduction Ecosystem services (ESS) are of high interest for ecological research and for decision-making in biodiversity conservation and management (MEA, 2005; Mace et al., 2012). Of the four types of ESS recognized by the MEA (2005), least attention has been paid toward cultural ecosystem services, such as nature-based tourism, outdoor recreation, landscape aesthetics, or cultural heritage (MEA, 2005; Cardinale et al., 2012). Among forms of nature-based tourism, wildlife tourism is particularly important as it has a great potential to provide revenues benefiting local populations and regional economies, while simultaneously supporting biodiversity conservation (Gössling, 1999; MEA, 2005;

⁎ Corresponding author at: Senckenberg Biodiversity and Climate Research Centre (BiKF), Senckenberganlage 25, 60325 Frankfurt, Germany. E-mail addresses: [email protected] (C. Grünewald), [email protected] (M. Schleuning), [email protected] (K. Böhning-Gaese).

http://dx.doi.org/10.1016/j.biocon.2016.05.036 0006-3207/© 2016 Elsevier Ltd All rights reserved.

Ballantyne et al., 2011; Buckley et al., 2012; Cimon-Morin et al., 2013). Wildlife tourist numbers and tourism revenues seem to be closely linked to tourist satisfaction and a destination's reputation (Gössling, 1999; Goodwin and Leader-Williams, 2000; Kerley et al., 2003). However, many destinations worldwide remain below their visitor as well as income and conservation funding capacities (Gössling, 1999; Balmford et al., 2015). Hence, management strategies are needed that meet tourist expectations and aim at sustainably increasing visitor numbers, while simultaneously conserving wildlife. Wildlife tourism can be directly linked to biodiversity (i.e. viewing or photographing wildlife; Goodwin and Leader-Williams, 2000; Mace et al., 2012), but might also be, directly or indirectly, influenced by a multitude of other factors, such as climate, ecosystem features (e.g. vegetation), landscape features (e.g. water systems) and infrastructure (e.g. Scott et al., 2004; Dramstad et al., 2006; Neuvonen et al., 2010). Consequently, management strategies for protected areas could impact wildlife tourism via biodiversity management, but also via modifications of ecosystem and landscape features or infrastructure (Goodwin and Leader-Williams, 2000;

C. Grünewald et al. / Biological Conservation 201 (2016) 60–68

Hanks, 2000; Peel et al., 2004; Neuvonen et al., 2010). Thus far, there is a lack of comprehensive quantitative studies aiming to understand how this multitude of factors and particularly biodiversity influence wildlife tourists and how these factors are related to tourist expectations. Large African savannah national parks are suitable systems to study the direct and indirect relationships between wildlife tourism and its multiple determinants and to simultaneously investigate tourist preferences. African savannah ecosystems harbor a unique diversity of large mammals, ungulates and large predators, and are important wildlife tourism destinations (Goodwin and Leader-Williams, 2000; Boshoff et al., 2007). Wildlife tourism is of high economic relevance for many African countries and many national parks depend on tourism (Goodwin and Leader-Williams, 2000). African savannah national parks have been studied in much detail, which allows formulating a priori hypotheses on the direct and indirect relationships between wildlife tourism and its determinants (Fig. 1): Large mammals, such as ungulates and predators, are expected to attract national park visitors (Goodwin and Leader-Williams, 2000; Lindsey et al., 2007; Okello et al., 2008; Maciejewski and Kerley, 2014a,b) and a positive relationship between large mammal densities and visitor numbers is expected, but has rarely been demonstrated (but see Naidoo et al., 2011). In particular, encountering specific animal species, such as charismatic species like lion or elephant, seems crucial for visitor satisfaction (Goodwin and Leader-Williams, 2000; Kerley et al., 2003). However, park visitors might also be attracted by open habitats with a good visibility (Goodwin and Leader-Williams, 2000; Peel et al., 2004), by scenic landscapes or by certain standards in visitor infrastructure (Goodwin and Leader-Williams, 2000; Turpie and Joubert, 2001; Beh and Bruyere, 2007; Lindsey et al., 2007). Given this interplay of different factors, park authorities could improve visitors' wildlife experience and potentially increase visitor numbers and economic benefits at multiple levels. First, management could control biodiversity directly (e.g. population regulation of ungulates and large predators), as well as indirectly, for example via vegetation cover and visibility (e.g. prescribed burning; Peel et al., 2004). Second, management could supply specific landscape features (e.g. surface water; Smit et al., 2007; Valeix et al., 2010) or invest in the establishment and maintenance of infrastructure (Goodwin and Leader-Williams, 2000; Hanks, 2000). To develop efficient management strategies and to sustain biodiversity and profitable wildlife tourism, a quantitative understanding of the relationships between wildlife tourism and its determinants is required. In addition, solid knowledge on wildlife tourists' expectations is needed. In this study, we investigate the role of multiple potential determinants of the distribution and behavior of wildlife tourists in a large African savannah national park with structural equation modeling (SEM). We additionally assess wildlife tourists' expectations and preferences

Fig. 1. Potential direct and indirect relationships (depicted in bold) between wildlife tourism and biodiversity, as well as infrastructure, ecosystem features (e.g. vegetation), visibility, landscape features, and climate. Biodiversity, in particular large mammals, might positively influence tourist numbers which could also be influenced by good visibility, scenic landscapes or high standards in infrastructure. Management could affect wildlife tourists' experiences at several levels, e.g. in regulating biodiversity or visibility and in supplying certain landscape features or in investing in infrastructure.

61

in interviews. To address these objectives, we simultaneously collected large mammal and visitor data along 78 road transects in Kruger National Park (KNP), South Africa, and conducted visitor interviews. We also quantified vegetation parameters and visibility, and compiled data on landscape features, infrastructure and climatic factors. We specifically analyze (i) the relationship between biodiversity and park visitors and test how visitors distribute across the park in relation to ungulates, large predators and specific charismatic species (lion, leopard, elephant, giraffe, buffalo). We validate these relationships by assessing visitors' wildlife viewing preferences derived from interviews. We predict that park visitors aggregate in areas of the park where sighting probabilities of animals are particularly high and that these visitor aggregations are driven by preferences for specific wildlife. (ii) We further test how other factors, such as vegetation cover, visibility, surface water (e.g. waterholes), infrastructure (e.g. proximity to main camps) and climate may directly and indirectly influence visitor numbers within the park. We then link these findings to visitors' perception of visibility and landscape features, and infrastructure preferences. The results from quantitative statistical models and interviews could inform park management to enhance tourist satisfaction, and thus increase tourist numbers and potentially tourism revenues.

2. Material and methods 2.1. Study area, road counts and visitor interviews Kruger National Park (KNP; South Africa) is one of the largest parks in Africa (~20,000 km2). It is well-managed, maintains large and relatively stable animal populations and is a main tourist attraction with approximately 1.5 million annual visitors (SANParks, 2014). The study was conducted in the southern part of KNP (~ 9000 km2; 24°57′1″ S, 31°36′20″ E) which covers different vegetation types (11 of 15 different types based on Jacana Media map [1999], provided by KNP Scientific Services as the most recent vegetation map) and a rainfall gradient from 350 to 850 mm (Fig. 2). To assess visitors' wildlife experience, we sampled our data along the public road network; these roads are accessible to all visitors. Within the public road network, we employed a stratified random approach to distribute road transects across the different gradients in rainfall and vegetation of southern KNP. Transects were approximately evenly distributed across the different classes of both rainfall (Fig. 2) and vegetation. Transects had a length of 5 km with a minimum distance of 1 km between transects and to the next main camp or gate, resulting in 78 road transects (Fig. 2). We replicated each transect three times in three consecutive months in the dry season (June, July, August 2012) and counted large mammals and visitors along the road transects (Fig. 2). Predators are elusive; we therefore combined our own predator sightings with information from visitors on their predator sightings from interviews (see below for details). We also collected information on visitors' expectations, park experience and wildlife viewing preferences in interviews. We developed a standardized questionnaire in collaboration with KNP and SANParks scientists that comprised four thematic sections: 1) general visitor details; 2) activities in the park; 3) park and game viewing experiences; 4) vegetation and visibility (see Appendix S2-A for full questionnaire). We used the questionnaire in face-to-face interviews with park visitors, conducted at all picnic sites and in all main camps that were located within our study area (five picnic sites and five main camps; Fig. 2). Interviews were conducted around lunch time as many visitors were available in public areas at this time of the day and willing to participate in interviews. We visited every interview spot three times, approximately once per month in June, July and August. Each time, we randomly selected six visitors per site as the maximum feasible number of interviewees during lunch time. In the two largest camps Skukuza and Lower Sabie, we randomly selected ten visitors, resulting in 204 interviews in total.

62

C. Grünewald et al. / Biological Conservation 201 (2016) 60–68

Germany). “Short grass” was defined as grass that was almost grazed to the ground. As shrubs and trees are difficult to distinguish (Zizka et al., 2014), we defined all woody plants b 3 m as shrubs and N3 m as trees. High visibility is probably closely related to visitors' wildlife experience as visitors might prefer open areas in which wildlife is easily spotted (e.g. Peel et al., 2004). To measure visibility in each transect, we estimated the distance at which we would be able to detect a fictitious adult warthog (Phacochoerus africanus) to the left and to the right hand side of the road; this approach is an established method for estimating visibility in African savannah (see Caro, 1999a,b). We calculated mean values of each variable for each transect (Table 1). 2.3. Biodiversity data and park visitors 2.3.1. Ungulate densities Ungulate densities were determined in standardized counts by driving along each transect at a constant, slow speed of 15 km/h in a 4WD pick-up truck (VW Amarok) with two observers at a time, one screening the right and the other the left side of the road. We conducted the study with a team of four observers (one “permanent” and three part-time observers). To minimize observer bias, all observers were well-trained (i.e. several days of training and comparisons of estimates between team members before transect counts), and we randomized driver and passenger positions in the car within a pair of observers. Additionally, we randomly allocated observer teams to transect identities and to transect

Table 1 List of all predictor variables considered in the analysis of ungulate densities, predator sighting probability and visitor vehicles. Predictor variables were divided into two thematic groups and five subgroups and given are the results, t-values and their significance level, and adjusted R2-values of simple linear regressions in which every variable was tested against ungulate densities, predator sighting probability and visitor vehicles, respectively. Predictor variables Fig. 2. Map of the study area in southern Kruger National Park (KNP) showing the locations of 78 road transects, the public road network, the park entry gates and the distribution of mean annual rainfall (different shadings, see legend). Also given are the locations of the visitor interview sites (5 main camps and 5 picnic sites).

2.2. Climate and ecosystem variables 2.2.1. Annual rainfall We focused on annual rainfall as climatic variable as rainfall varies substantially across southern KNP and is one of the most important climatic variables shaping African savannah systems including KNP (e.g. Gaylard et al., 2003; Sankaran et al., 2005; Shorrocks, 2007). Temperature varies only slightly within southern KNP and was therefore not included in the study. We considered mean annual rainfall per transect for the year of study (2012). Rainfall values for each transect were derived in a geographic information system (ARCGIS v. 10.1) from high-resolution maps on rainfall provided by KNP Scientific Services. The precipitation data used for our study, collected at 22 weather stations in southern KNP, have an excellent spatial resolution and the maps on annual rainfall extrapolated from the local weather stations can be precisely matched to the scale of our transects.

2.2.2. Vegetation cover and visibility We quantified vegetation cover and visibility along each transect at every kilometer (at 0, 1, 2, 3, 4 and 5 km) from the vehicle as this reflects the visitors' perspective. We recorded vegetation cover and visibility once during our field season as we observed little change over the three months. We quantified vegetation cover as the proportion of grass (short, long grass), shrub, tree, bare ground, burned area and rock in squares of 20 × 20 m to the left and right of the road. We estimated plot dimensions with a laser range finder (Nikon Laser800S, Nikon,

Simple linear regression Ungulates t

I. Climate & ecosystem 1. Climate Mean annual rainfall 2. Vegetation % long grass % short grass (grazed to ground) % shrub (woody plant b 3 m) % tree (woody plant N 3 m) % bare ground % burned area % rock

Predators 2

R

Vehicles 2

t

R

−3.35⁎⁎

0.12

0.77

b0.01

−3.07⁎⁎ 5.65⁎⁎⁎

0.10 0.29

0.03 1.01

b0.01 b0.01

1.13

b0.01

0.83

b0.01

−0.18

b0.01

0.45

b0.01

3.58⁎⁎⁎ −2.59⁎ −1.20

0.13 −3.23⁎⁎ 0.07 −0.58 0.01 0.89

R2

t

0.11 b0.01 b0.01

II. Landscape & infrastructure 3. Surface water Proximity to river Proximity to waterhole 4. Visibility Distance to warthog 5. Infrastructure Proximity to gate Proximity to main camp Proximity to picnic site Proximity to hide Proximity to get-off point

2.72⁎⁎ 1.21

0.08 0.01

1.94 1.24

1.26

0.01

0.50

Significant relationships are given in bold. ⁎ p b 0.05. ⁎⁎ p b 0.01. ⁎⁎⁎ p b 0.001.

0.04 2.74⁎⁎ 0.08 0.01 −0.57 b0.01

b0.01 −0.94

b0.01

1.90 0.37

0.03 b0.01

−1.90

0.03

0.60 2.30⁎

b0.01 0.05

C. Grünewald et al. / Biological Conservation 201 (2016) 60–68

replicates. We replicated each transect three times and counted all large mammals (e.g. zebras (Equus quagga), giraffes (Giraffa camelopardalis), antelopes, elephants (Loxodonta africana); see Appendix S1, Table S1 for species lists). We excluded hippos (Hippopotamus amphibius), baboons (Papio spp.) and other monkeys because their habitat and locomotion differ greatly from other large African mammals. We followed well-established standard protocols for distance sampling to avoid sampling bias (Buckland et al., 2001). When an animal was spotted, we identified the species, recorded the number of individuals and measured the distance from the observer to the individual (or the center of the herd) with a range finder and measured the angle between the center line of the road transect and the individual with a compass. In the rare case of fast moving animals, we assessed these measures for the first spotting location of the respective animals. To further reduce sampling bias on transects, we sampled transects in a random order and we sampled each transect at different visits, in different driving directions (start to end, end to start) and during different daytimes, i.e. in the early morning (6:45 a.m. to 9:30 a.m.), late morning (9:45 a.m. to 12:15 p.m.) and in the afternoon (1:30 p.m. to 4:45 p.m.). From these counts, we calculated ungulate densities using DISTANCE v. 6.0 (Thomas et al., 2009). DISTANCE takes into account detection probabilities of the observed ungulates which decrease with increasing distance from the line of observation. We calculated ungulate densities by fitting a single detection function across all ungulate species and all three counts and transects (n = 234). We chose the best function and adjustment terms based on AIC-values. Ungulate densities were averaged over the three counts per transect (n = 78). Since we averaged data over the three replicates, we minimize observer bias and effects of other confounding factors, ensuring that the data adequately represent wildlife spotting opportunities for visitors in KNP during the study period.

2.3.2. Predator sighting probability To estimate how likely it was for park visitors to see large predators along the transects, we used own predator sightings as well as data from visitor interviews. We considered lion (Panthera leo), leopard (Panthera pardus), spotted hyena (Crocuta crocuta), cheetah (Acinonyx jubatus) and wild dog (Lycaon pictus) because they are the largest predators in KNP, are charismatic and prey on a variety of ungulate species (Shorrocks, 2007). Data on predator sightings were collected during the standardized ungulate counts using the same methodology as for ungulates. Due to the elusiveness and rarity of most predator species, transect counts generated rather little predator data. Therefore, we recorded predator sightings also during other drives in the park. We extracted all predator sightings that fell within one of our transects and quantified the number of predator sightings for each transect. We also tracked all our routes in the park with a GPS-device and assessed the number of times we had driven each transect. To supplement our predator data, we conducted interviews, which are commonly used in predator assessments (Fanshawe et al., 1991; Gros et al., 1996; Gros, 2002). We asked park visitors during the interviews to mark predator sightings on a map and also to mark the routes they had driven on the day of interview. We digitized this information (ARCGIS v. 10.1) and again extracted all sightings that were located within one of our transects and quantified the number of predator sightings for each transect. From the routes provided in the interviews, we also assessed the number of times each transect had been driven by visitors. With data from these different, but comparable sources of information, we calculated sighting probabilities of predators by dividing the number of predator sightings by the number of drives for each transect. We calculated both a predator sighting probability across all five species and individual sighting probabilities for each predator species. Sighting probabilities of predators by visitors are most suitable for our purpose as they directly reflect the visitors' perspective and assess the likelihood of a predator sighting for a given transect.

63

2.3.3. Visitor numbers To assess visitor numbers for each transect, we counted the number of visitor vehicles and the number of passengers per vehicle passing us in the opposite direction during the standardized ungulate counts. Thus, visitor counts were also replicated three times per transect and we determined the number of vehicles, the number of passengers and the time period that was needed to cover each replicate. From these data we calculated the number of vehicles and the number of passengers per hour. As visitor and vehicle numbers in the park varied over time, e.g. visitor numbers were higher on weekends, we corrected the visitor counts per transect for overall visitor and vehicle numbers in the park. Data on overall numbers of vehicles and visitors in southern KNP were derived from daily gate entry statistics of all gates bordering our study area (Fig. 2; provided by KNP Scientific Services). We related for each day of counting (n = 234), the numbers of passengers/h and vehicles/h from the transect counts to the number of visitors and vehicles in the study area, respectively. To meet model assumptions of linear regressions, we square root-transformed vehicles/h and passengers/h, obtained the residuals of the linear regressions and averaged them over the three temporal replicates per transect (n = 78). The residuals of the no. of vehicles and no. of passengers were highly correlated (rp = 0.99, n = 78) and therefore all subsequent statistical analyses were conducted with the residuals of the no. of vehicles. 2.4. Landscape and infrastructure variables To quantify landscape and infrastructure features, we measured the minimum distance from the center of each transect to the nearest waterhole and to the nearest main river in a geographic information system (ARCGIS v. 10.1; landscape maps provided by KNP Scientific Services). We used the very latest information on waterholes that was available and only included waterholes that were classified as open and that were accessible to visitors throughout the study period. We also measured the minimum distance of each transect to important infrastructural features, e.g. the next entry gate, main camp, picnic site, hide and get-off point (i.e. areas in the park in which visitors are allowed to get off their vehicles). Both get-off points and hides are usually scenic look-outs and/or good wildlife viewing locations. All measurements were based on KNP digital infrastructure maps. For statistical analyses, we transformed all distance measurements into proximities by multiplying these variables with minus one. 2.5. Statistical analysis 2.5.1. Variable selection We tested the direct and indirect links between climate, vegetation, visibility, landscape features, large mammals, infrastructure and visitors with structural equation modeling (SEM). SEMs allow testing a priori hypotheses of such links and, despite their correlative nature, SEMs are powerful tools for investigating complex ecological relationships at large, non-experimental spatial scales (Anderson et al., 2010; Grace et al., 2012). With SEMs, total effects (correlations between predictor and response variables) can be partitioned into direct and indirect effects (links via other predictor variables) revealing mechanisms that might remain undetected in multiple regressions (Grace et al., 2012). To reduce problems of collinearity, we performed a variable selection prior to SEM fitting to identify the most important variables. To improve normality of residuals, ungulate densities were square root-transformed, all vegetation cover data were arc sine-square root-transformed and visibility and probability of predator sightings were logtransformed. First, all predictor variables were grouped into two thematic groups and five subgroups (Table 1): I. Climate & ecosystem variables: 1. climate (i.e. annual rainfall); 2. vegetation (e.g. grass cover); II. Landscape & infrastructure: 3. surface water (e.g. proximity to river); 4. visibility; 5.

64

C. Grünewald et al. / Biological Conservation 201 (2016) 60–68

infrastructure (e.g. proximity to gate, Table 1). Second, we tested each of the initial predictor variables in simple linear regressions against ungulate densities, predator sighting probability and no. of vehicles if a logical ecological or socio-economic relationship was assumed (Fig. 1, Table 1). Only relationships with p-values b 0.05 in simple regressions were considered in the further selection process (Table 1). Third, within the five subgroups, we selected the most important variable predicting ungulate densities, predator sighting probability and no. of vehicles, based on the adjusted R2-values of the simple linear regressions. In this procedure, mean annual rainfall, percentage cover of short grass and proximity to rivers were selected in relation to ungulate densities, and percentage cover of bare ground and proximities to rivers and to getoff points were selected in relation to predators and to no. of visitor vehicles, respectively (Table 1). All statistical analyses in the pre-selection process were performed with R 2.11.1 (R Core Team, 2010). 2.5.2. Structural equation modeling (SEM) We developed an a priori SEM (Fig. 3) based on the statistical variable pre-selection and based on the conceptual relationships in Fig. 1. The a priori SEM was tested in AMOS, an SPSS extension (version 19.0.0). We evaluated each SEM utilizing common measures and indices of model fit, e.g. the comparative fit index CFI (CFI N 0.90 indicates a good fit) and chi-square tests (p N 0.05 indicates a good model; see Grace et al., 2012). We allowed correlations amongst all predictor variables in the model (mean annual rainfall, proximities to rivers and to get-off points). A balance between model complexity and sample size is required to avoid model misinterpretation due to unreliable coefficient estimates (Grace et al., 2012). To this end, we further simplified the a priori SEM and stepwise removed all non-significant paths in the SEM (starting with the least significant) until all paths in the model were at least marginally significant (p-value b 0.1; Grace et al., 2012). In addition to testing the final SEM by itself, it also served as a template to test more detailed hypotheses. First, we calculated SEMs for predator sighting probabilities of every single predator species, linking them to overall ungulate densities and to the no. of vehicles. Second, we tested the importance of specific charismatic ungulate species for no. of vehicles. We therefore calculated the densities of elephant, rhino (i.e. Ceratotherium simum), zebra, giraffe and buffalo (Syncerus caffer) in DISTANCE v. 6.0 (Thomas et al., 2009). We employed the same detection function and adjustments as for overall ungulate densities, but obtained density estimates per transect for individual species. We again averaged densities from the three counts per species and transect. We then replaced overall ungulate densities by the densities of the selected charismatic species in the final SEM and included an additional path between the respective species and no. of vehicles. We calculated a separate SEM for each of the five predator and five ungulate species (see Appendix S3).

Fig. 3. A priori structural equation model (SEM) with direct and indirect links between climatic factors (i.e. annual rainfall), proximity to rivers and get-off points, percentage cover of bare ground and short grass, ungulate densities, predator sightings and wildlife tourism (i.e. no. of visitor vehicles); the a priori SEM is based on a statistical variable selection process (see 2. Materials and methods, 2.5 Statistical analysis).

We tested the statistical models for potential spatial autocorrelation by calculating Moran's I (spdep and ncf package, permutation approach with 1000 permutations; R Core Team, 2010) of the residuals of multiple regressions corresponding to the final SEM (Fig. 4). We calculated regressions for each endogenous variable in the final SEM as the response variable, including all variables with a direct link to the respective endogenous variable as predictors (Kissling et al., 2008). We did not detect significant spatial autocorrelation in the residuals of any model (see Appendix S4). 3. Results 3.1. Final SEM The final SEM (Fig. 4, Table 2) yielded good measures of fit and identified a significant positive relationship between biodiversity and visitor numbers. In particular, large predator sighting probability and no. of vehicles were positively associated, i.e. more visitor vehicles were found in transects where the probability of seeing a large predator was high (Fig. 4). When fitting the final SEM with individual predator species, we detected stronger relationships between lion and leopard sighting probabilities and no. of vehicles compared to the other predator species (see Appendix S3, Table S3). Ungulate densities were not directly associated with no. of vehicles. Moreover, none of the charismatic ungulate species, i.e. elephant, rhino, giraffe, buffalo, were directly related to no. of visitor vehicles (see Appendix S3, Table S2). However, ungulate densities were indirectly related to no. of vehicles via the large predators (Fig. 4, Table 2). Our model revealed direct and indirect relationships between several non-biodiversity factors and visitors. Vehicle numbers were higher in transects that were located close to rivers and to get-off points (Fig. 4), whereas transects in close proximity to waterholes and with a good visibility were not related to high vehicle numbers (Table 1). Proximity of transects to rivers, as well as vegetation cover in the transects (i.e. amount of bare ground and of short grass) were also indirectly associated with no. of vehicles via ungulate densities and large predators (Fig. 4, Table 2). Finally, annual rainfall showed complex patterns and was negatively related to ungulate densities (Fig. 4, Table 2), but weakly and positively associated with predators (Table 2). The models additionally showed an indirect, but weak relationship between rainfall and visitor numbers (Table 2). 3.2. Visitor expectations and wildlife viewing preferences The park visitors in our interviews were mainly locals (79% South African, n = 204 visitors) and rather experienced (N60% regular visitors, n = 204; see Appendix S2-B for visitor details). When assessing wildlife viewing experiences and visitor expectations, we only included interviewees that had spent the day of the interview on a drive in the park (n = 196). The majority of interviewees (65%, n = 196) visited the park to see specific mammal species, especially large predators (see Appendix S2-B), while other factors, such as family and relaxation time (4%) or infrastructural features (0%) were less important. When asked for the most important animal species that they wanted to see (“must see”; Fig. 5), visitors ranked lion and leopard highest, followed by cheetah. Among species they “expected to see” during their stay, visitors ranked elephants highest, followed by impala (Aepyceros melampus) and rhino (Fig. 5; see also Appendix S2-B). Eighty-three percent of the visitors (n = 196) perceived wildlife viewing as easy, mainly because of their spotting experience (18%) and the open winter vegetation (17%). Forty-eight percent of visitors (n = 196) felt that vegetation did not impair visibility and wildlife viewing. A majority of visitors (72%; n = 196) further described vegetation as positive for their park experience because they considered it an important part of nature and landscape (20%), they enjoyed looking for specific vegetation (14%) or recognized the value of animal-vegetation-interactions (13%). The

C. Grünewald et al. / Biological Conservation 201 (2016) 60–68

65

Fig. 4. Structural equation model (SEM) directly and indirectly linking climatic factors (annual rainfall), proximity to rivers and get-off points, percentage cover of short grass and bare ground with ungulate densities, predator sighting probability and no. of visitor vehicles. The SEM was developed based on the a priori SEM (Fig. 3) by statistically removing redundant, non-significant paths (if p N 0.1). Illustrated are direct effects (i.e. standardized partial regression coefficients) and their significance level for all significant paths (*p b 0.05; ***p b 0.001) or marginally significant paths (+ p b 0.1) and R2-values of all endogenous variables. Arrow width corresponds to effect size; positive relationships are shown with solid arrows, negative relationships with dashed arrows. The SEM yielded good measures of fit (CFI = 0.95; Chi2 = 17.66, df = 11, p = 0.09).

variety of landscapes, as well as savannah systems, and savannah tree species ranked highest among favorite landscapes, followed by water systems (e.g. rivers; see Appendix S2-B for more detailed result descriptions).

sightings among visitors and by visitor experience. Similar visitor behavior is found in parks and reserves in Kenya and South Africa where visitors spend large proportions of their viewing time on large predators such as lion, leopard and cheetah (Okello et al., 2008; Maciejewski and Kerley, 2014b). These results are also corroborated by our interviews that showed distinct wildlife viewing preferences of KNP visitors. Visitors preferred lions and leopards over cheetahs, wild dogs and spotted hyenas and generally ranked predators higher in importance than ungulate species. Seeing large mammals, in particular lions and leopards, was additionally the most frequently stated reason for visiting KNP. The weak relationship between park visitors and wild dogs might be due to the rarity of this species or a lack of visitors' awareness on their endangered status (e.g. Di Minin et al., 2012). Visitors of multiple other parks and reserves in southern Africa, Tanzania and Zambia have stated similar species preferences and rank charismatic wildlife among the most important reasons for their park visits (e.g. Goodwin and Leader-Williams, 2000; Boshoff et al., 2007; Lindsey et al., 2007; Di Minin et al., 2012). Here we show that visitors' preferences for specific wildlife is indeed detectable in quantitative analyses across road transects in KNP. We found that areas of high predator sighting probabilities in KNP were also associated with areas of high ungulate densities. However, visitor numbers along the transects were neither directly related to high ungulate densities nor to densities of charismatic ungulate species (see Appendix S3, Table S2). This finding is in line with previous findings from other African savannah national parks (Goodwin and Leader-Williams, 2000; Okello et al., 2008; Di Minin et al., 2012;

4. Discussion We tested the relationships between wildlife tourism and biodiversity along road transects in KNP and demonstrate that large predators, in particular lions and leopards, attract park visitors to certain areas of the park. Moreover, landscape features and infrastructure were positively associated with visitors, since visitor numbers were higher in transects close to rivers and get-off points. Additionally, an interplay of multiple direct and indirect factors, including ungulate densities, ecosystem and landscape features and climate, influenced predator sighting probability and visitor numbers. The relationships found in our correlative models were largely consistent with the expectations and preferences of park visitors revealed by our interviews. By combining these two methodological approaches, this is one of the first studies to show that preferences of wildlife tourists directly influence their behavior and distribution in a large African national park. Different species and guilds of species influenced visitor numbers differently. More visitor vehicles were found in areas with high sighting probabilities of lions and leopards, whereas sighting probabilities of spotted hyenas and wild dogs were not associated with high visitor numbers (see Appendix S3, Table S3). This association between specific wildlife and visitor numbers could be driven by the exchange of

Table 2 Standardized direct, indirect and total effects of annual rainfall, proximity to rivers, vegetation and proximity to get-off points on ungulate densities, predator sighting probability and no. of vehicles derived from the SEM in Fig. 4. Ungulates

Predators

Vehicles

Predictor

Direct

Indirect

Total

Direct

Indirect

Total

Direct

Indirect

Total

Mean annual rainfall Proximity to river % short grass % bare ground Proximity to get-off points Ungulates Predators

−0.24 n.a. 0.48

−0.10 0.16 n.a.

−0.35 0.16 0.48

n.a. 0.18 n.a. −0.47

0.08 0.05 0.15 n.a.

0.08 0.22 0.15 −0.47

0.30

n.a.

n.a. 0.23 n.a. n.a. 0.19 n.a. 0.30

0.02 0.07 0.04 −0.14 n.a. 0.09 n.a.

0.02 0.30 0.04 −0.14 0.19 0.09 0.30

n.a. = not applicable.

0.30

66

C. Grünewald et al. / Biological Conservation 201 (2016) 60–68

Fig. 5. Wildlife viewing preferences of KNP park visitors. Wildlife tourists were asked for the most important species (or groups of species) they wanted to see (“must see”) and “expected to see” while being in the park; given are the Top 5 species that ranked highest in each category (lion was in the Top 5 of both categories; see Appendix S2-B for top 10 lists).

Maciejewski and Kerley, 2014a). For example in Addo Elephant NP, elephant densities are not linked to annual numbers of park visitors (Maciejewski and Kerley, 2014a). Similarly, in Amboseli NP, visitors stop less frequently at elephant sightings and pay hardly any attention to buffalos (Okello et al., 2008). This suggests that visitor behavior and distributions along transects are primarily driven by visitors' preferences for animal species other than ungulates, i.e. large predators. Nevertheless, elephants and other large charismatic ungulates were also appreciated by KNP visitors in our interviews. In contrast to large predators, visitors largely expected to see these ungulate species and thus may not have actively searched for these species. Hence, it seems that large predators are primarily driving visitors' behavior, while elephants and other large charismatic ungulates are perceived as attractive, but probably rather as “add-ons”. In other African savannah parks, large charismatic ungulates are also appreciated, but less sought after than lions and other big cats (e.g. Lindsey et al., 2007; Okello et al., 2008; Di Minin et al., 2012). However, in the case of charismatic ungulate species, behavioral patterns of visitors might be related to species abundance and could differ among parks. If a species is particularly rare in a park, it might become more interesting for visitors (e.g. giraffes in Amboseli NP; Goodwin and Leader-Williams, 2000; Kerley et al., 2003; Okello et al., 2008; Maciejewski and Kerley, 2014b). Our results have important implications for management (Fig. 1). Three main levels can be identified as relevant for management, i.e. vegetation and surface water, ungulate populations and predators. These components potentially influence each other via bottom-up and topdown effects and interact with management actions (Shorrocks, 2007). The results of our statistical models and the interviews in KNP identify large predators as most attractive for visitors and most important for influencing visitors' behavior. This suggests that park management should primarily aim at high predator sighting probabilities to increase visitor satisfaction. One management option could be to introduce predators or to increase their numbers. Predator introductions can be successful and ecologically sensible in some park systems, but have also been discussed critically as they might result in multiple negative feedbacks (Shorrocks, 2007; Ripple et al., 2014). Our models demonstrate that high probabilities of predator sightings were related to high ungulate densities and indirectly to high proportions of short grass, surface water and annual rainfall. Thus, predator populations are closely associated with prey densities and other components of the savannah ecosystem. Hence, the restocking of ungulate populations could be a second management option which might however have both positive and negative effects on predators as well as on other components of the ecosystem (Shorrocks, 2007; Maciejewski and Kerley, 2014b). Our results further show that predators link vegetation, surface water and ungulate populations with wildlife tourism (Fig. 4). This suggests that

the above-mentioned one-dimensional management strategies of predator or ungulate introductions and restocking may not be sufficient. Instead, our results support an ecosystem-based management approach that involves management not only for predators, but also for ungulate populations, predator-prey interactions and specific vegetation structures including the multiple feedbacks between these different levels. With such an ecosystem-based management approach, predators could be sustained at high numbers which contributes to the attraction of park visitors. Another important management option to increase visitor satisfaction could involve better visitor information and exchange (e.g. via sighting boards) which may enhance the probability of predator sightings (Beh and Bruyere, 2007; Okello et al., 2008; Di Minin et al., 2012). Our results support the official SANParks (2006) policies for managing South Africa's national parks. Here predator management and introductions are considered as options for improving the ecotourism value of parks. However, this option is promoted only in the context of other ecological processes, should be based on scientific evidence and minimize its potential negative impacts. Additionally, the promotion of South Africa's biodiversity heritage and visitor education are considered important to enhance the attractiveness of wildlife tourism (SANParks, 2006). Besides large predators, only a few other factors were directly associated with visitor’ distribution across the park. Surprisingly, visibility along the road transects in KNP was not related to the number of visitors. Previous studies suggest that increased vegetation density and decreased visibility in African savannah parks result in a reduced wildlife experience (Goodwin and Leader-Williams, 2000; Kerley et al., 2003; Peel et al., 2004). Visitors of Addo Elephant NP, for example, have suggested that cutting down the dense natural thicket in the park would improve wildlife viewing (Kerley et al., 2003). Different to northern KNP or Addo Elephant NP, southern KNP lacks very dense vegetation, especially during the dry season. Nonetheless, southern KNP covers a variety of vegetation types and our transects were distributed across this vegetation gradient from very open (open savannah) to densely vegetated (thornbush thickets). Moreover, the results of the statistical models were in line with visitor expectations. KNP visitors felt that vegetation did not impair wildlife viewing. In contrast, vegetation was perceived as an important part of nature and scenery that positively contributed to wildlife experience. Accordingly, our model and interview results suggest that managing for artificially open vegetation will not increase visitors' satisfaction and that vegetation and visibility should not be artificially controlled (e.g. by prescribed burning; Peel et al., 2004). However, the scientific evidence for the link between vegetation, visibility and wildlife tourism remains scarce and further investigations might be helpful to determine if our results hold in different seasons and across different parks. Nonetheless, some studies from Kenyan parks corroborate management strategies that favor the promotion of natural vegetation and wilderness to improve visitors' wildlife experience (Beh and Bruyere, 2007; Okello et al., 2008). Visitor numbers in KNP were positively associated with short distances to rivers. Visitors of KNP and of African parks in general appear to appreciate rivers in their aesthetic value and as part of the scenery (Goodwin and Leader-Williams, 2000; Turpie and Joubert, 2001; Lindsey et al., 2007). Moreover, rivers are known as good spots for observing wildlife, which was also reflected by an indirect positive link between rivers and visitors in KNP mediated via vegetation structure, ungulates and large predators (Fig. 4). This is consistent with previous studies showing that rivers in African national parks act as important water resources, but also provide crucial places for resting, hunting and feeding of mammals. Thus, visitors are likely to observe animal behavior which is highly attractive for visitors (Hopcraft et al., 2005; Smit et al., 2007; Okello et al., 2008). Different to previous studies (Smit et al., 2007; Valeix et al., 2010), we did not find a significant direct nor indirect relationship between waterholes and large mammals or visitor numbers, indicating that natural rivers were more important for driving animal and visitor behavior than artificial waterholes in KNP (see also

C. Grünewald et al. / Biological Conservation 201 (2016) 60–68

Turpie and Joubert, 2001). High numbers of visitor vehicles in transects close to get-off points might also be driven by aesthetic factors. Get-off points in KNP are scenic look-outs, often located upon hills and near rivers or dams, and thus offer good wildlife viewing opportunities. These results were again congruent with visitor expectations as KNP visitors ranked scenic and diverse landscapes highest and also appreciated specific landscape types, e.g. water systems. In Kenyan parks, scenic view points and aesthetic landscapes are also highlighted as important for visitor satisfaction (Beh and Bruyere, 2007; Okello et al., 2008). Unexpectedly, we did not find a link between other infrastructure features (e.g. proximity to main camp, gate) and visitors' distribution along transects. Consistently, infrastructure did not shape visitors' expectations in KNP. Infrastructure features might be less important for visitors in KNP than in other parks as KNP has a good road network and well-maintained infrastructure, which is not the case in all African parks (Goodwin and Leader-Williams, 2000; Hanks, 2000). In summary, our findings from both models and interviews suggest that park management ought to preserve natural structures and habitats to increase wildlife and park experience rather than further developing artificial park features and infrastructure. In general, natural, authentic as well as scenic landscapes positively contribute to wildlife experience in African savannah parks, whereas comfort and luxury seem to be of lower importance (Goodwin and Leader-Williams, 2000; Beh and Bruyere, 2007; Lindsey et al., 2007; Okello et al., 2008). Biodiversity, especially large predators, and non-biodiversity components, such as natural landscape features, rank high in park visitors' expectations and influence visitor behavior and distribution across a large African savannah national park. This indicates that conserving near-natural ecosystems with diverse wildlife is essential for visitor satisfaction and park attractiveness for wildlife tourism in KNP as well as in other African savannah national parks. Park management should therefore promote natural features and focus on conservation and restoration strategies instead of investing in new infrastructure. Such an approach could be supplemented by ecotourism-focused marketing and may also maximize the economic revenues and the capacities for conservation funding in African national parks. We conclude that biodiversity conservation should focus on the maintenance of near-natural savannah ecosystems, while simultaneously providing attractive wildlife experiences to tourists. Such an ecosystem-based management approach may help to develop African national parks towards a sustainable and profitable wildlife tourism in the long run. Funding sources C. Grünewald was funded by a scholarship of the Cusanuswerk; additional funding was provided by LOEWE BiK-F, Hesse and the DFG (BO1221/19-1). Acknowledgments We thank M. Templin for field assistance and support throughout the project, L. Birkmann for field assistance and data processing, J. Wegfahrt for field assistance; S. Higgins for project facilitation and advice; SANParks and KNP for research permits; KNP Scientific Services for support of the project and help in the field, in particular M. Herbst, R. Scholtz and S. MacFadyen, but also N. Zambatis, D. Biggs and R. Grant; P. Khoza and the KNP Research Camp & Staff; M. Brummer and J. Zenner for data processing; the reviewers for their helpful comments; D. Bowler for proofreading and language editing; M. Niggemann and E. L. Neuschulz for advice. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.biocon.2016.05.036.

67

References Anderson, T.M., Hopcraft, J.G.C., Eby, S., Ritchie, M., Grace, J.B., Olff, H., 2010. Landscapescale analyses suggest both nutrient and antipredator advantages to Serengeti herbivore hotspots. Ecology 91, 1519–1529. Ballantyne, R., Packer, J., Sutherland, L.A., 2011. Visitors' memories of wildlife tourism: implications for the design of powerful interpretive experiences. Tour. Manag. 32, 770–779. Balmford, A., Green, J.M.H., Anderson, M., Beresford, J., Huang, C., Naidoo, R., Walpole, M., Manica, A., 2015. Walk on the wild side: estimating the global magnitude of visits to protected areas. PLoS Biol. 13. http://dx.doi.org/10.1371/journal.pbio.1002074. Beh, A., Bruyere, B.L., 2007. Segmentation by visitor motivation in three Kenyan national reserves. Tour. Manag. 28, 1464–1471. Boshoff, A.F., Landman, M., Kerley, G.I.H., Bradfield, M., 2007. Profiles, views and observations of visitors to the Addo Elephant National Park, Eastern Cape, South Africa. S. Afr. J. Wildl. Res. 37, 189–196. Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake, J.L., Borchers, D.L., Thomas, L., 2001. Introduction to Distance Sampling – Estimating Abundance of Biological Populations. Oxford University Press, Oxford. Buckley, R.C., Castley, J.G., de Vasconcellos Pegas, F., Mossaz, A.C., Steven, R., 2012. A population accounting approach to assess tourism contributions to conservation of IUCNredlisted mammal species. PLoS One 7, 1–8. Cardinale, B.J., et al., 2012. Biodiversity loss and its impact on humanity. Nature 486, 59–67. Caro, T.M., 1999a. Abundance and distribution of mammals in Katavi National Park, Tanzania. Afr. J. Ecol. 37, 305–313. Caro, T.M., 1999b. Densities of mammals in partially protected areas: the Katavi ecosystem of western Tanzania. J. Appl. Ecol. 36, 215–217. Cimon-Morin, J., Darveau, M., Poulin, M., 2013. Fostering synergies between ecosystem services and biodiversity in conservation planning: a review. Biol. Conserv. 166, 144–154. Di Minin, E., Fraser, I., Slotow, R., MacMillan, D.C., 2012. Understanding heterogeneous preference of tourists for big game species: implications for conservation and management. Anim. Conserv. 16, 249–258. Dramstad, W.E., Sundili Tveit, M., Fjellstad, W.J., Fry, G.L.A., 2006. Relationships between visual landscape preferences and map-based indicators of landscape structure. Landsc. Urban Plan. 78, 465–474. Fanshawe, J.H., Frame, L.H., Ginsberg, J.R., 1991. The wild dog — Africa's vanishing carnivore. Oryx 25, 137–146. Gaylard, A., Owen-Smith, N., Redfern, J., 2003. Surface water availability: implications for heterogeneity and ecosystem processes. In: du Toit, J.T., Rogers, K.H., Biggs, H.C. (Eds.), The Kruger Experience: Ecology and Management of Savanna Heterogeneity. Island Press, Washington, pp. 171–188. Goodwin, H.J., Leader-Williams, N., 2000. Tourism and protected areas – distorting conservation priorities towards charismatic megafauna? In: Entwistle, A., Dunstone, N. (Eds.), Priorities for the Conservation of Mammalian Diversity. Has the Panda Had Its Days?Cambridge University Press, Cambridge, pp. 257–275 Gössling, S., 1999. Ecotourism: a means to safeguard biodiversity and ecosystem functions? Ecol. Econ. 29, 302–320. Grace, J.B., Schoolmaster Jr., D.R., Guntenspergen, G.R., Little, A.M., Mitchell, B.R., Miller, K.M., Schweiger, E.W., 2012. Guidelines for a graph-theoretic implementation of structural equation models. Ecosphere 3 (Article 73). Gros, P.M., 2002. The status and conservation of the cheetah Acinonyx jubatus in Tanzania. Biol. Conserv. 106, 177–185. Gros, P.M., Kelly, M.J., Caro, T.M., 1996. Estimating carnivore densities for conservation purposes: indirect methods compared to baseline demographic data. Oikos 77, 197–206. Hanks, J., 2000. The role of transfrontier conservation areas in southern Africa in theconservation of mammalian biodiversity. In: Entwistle, A., Dunstone, N. (Eds.), Priorities for the Conservation of Mammalian Diversity. Has the Panda Had Its Days?Cambridge University Press, Cambridge, pp. 239–256. Hopcraft, J.G.C., Sinclair, A.R.E., Packer, C., 2005. Planning for success: Serengeti lions seek prey accessibility rather than abundance. J. Anim. Ecol. 74, 559–566. Kerley, G.I.H., Geach, B.G.S., Vial, C., 2003. Jumbos or bust: do tourists' perceptions lead to an under-appreciation of biodiversity? S. Afr. J. Wildl. Res. 33, 13–21. Kissling, W.D., Field, R., Böhning-Gaese, K., 2008. Spatial patterns of woody plant and bird diversity: functional relationships or environmental effects? Glob. Ecol. Biogeogr. 17, 327–339. Lindsey, P.A., Alexander, R., Mills, M.G.L., Romaňach, S., Woodroffe, R., 2007. Wildlife viewing preferences of visitors to protected areas in South Africa: implications for the role of ecotourism in conservation. J. Ecotourism 6, 19–32. Mace, G.M., Norris, K., Fitter, A.H., 2012. Biodiversity and ecosystem services: a multilayered relationship. Trends Ecol. Evol. 27, 19–26. Maciejewski, K., Kerley, G.I.H., 2014a. Elevated elephant density does not improve ecotourism opportunities: convergence in social and ecological objectives. Ecol. Appl. 24, 920–926. Maciejewski, K., Kerley, G.I.H., 2014b. Understanding tourists' preference for mammal species in private protected areas: is there a case for extralimital species for ecotourism? PLoS One 9, e88192. Millennium Ecosystem Assessment, 2005. Ecosystems and Human Well-being: Synthesis and Biodiversity Synthesis. World Resources Institute, Washington DC. Naidoo, R., Weaver, L.C., Stuart-Hill, G., Tagg, J., 2011. Effect of biodiversity on economic benefits from communal lands in Namibia. J. Appl. Ecol. 48, 310–316. Neuvonen, M., Pouta, E., Puustinen, J., Sievanen, T., 2010. Visits to national parks: effects of park characteristics and spatial demand. J. Nat. Conserv. 18, 224–229.

68

C. Grünewald et al. / Biological Conservation 201 (2016) 60–68

Okello, M.M., Manka, S.G., D'Amour, D.E., 2008. The relative importance of large mammal species for tourism in Amboseli National Park, Kenya. Tour. Manag. 29, 751–760. Peel, M.J.S., Davies, R., Hurt, R., 2004. The value and costs of wildlife and wildlife-based activities in the eastern Lowveld savannah, South Africa. In: Lawes, M.J.L., Ealey, H.A.C., Shackleton, C.M., Geach, B.G.S. (Eds.), Indigenous Forests and Woodlands in South Africa. University of KwaZulu-Natal Press, pp. 775–795. R Development Core Team, 2010. R, a Language and Environment for Statistical Computing. R foundation for Statistical Computing, Vienna, Austria (online: http://www.Rproject.org). Ripple, W.J., et al., 2014. Status and ecological effects of the world's largest carnivores. Science 343, 1241484. Sankaran, M., et al., 2005. Determinants of woody cover in African savannas. Nature 438, 846–849. SANParks, 2006. Coordinated Policy Framework Governing Park Management Plans (available online at: https://www.sanparks.org/docs/conservation/cpfjanuary2010. pdf) SANParks, (South African National Parks), South Africa (downloaded 7th of May 2016). SANParks, 2014. Annual Report 2013/2014 (available online at: www.sanparks.org/ assets/docs/general/annual-report-2014.pdf) SANParks, (South African National Parks), South Africa (downloaded 12th of February 2015).

Scott, D., McBoyle, G., Schwartzentruber, M., 2004. Climate change and the distribution of climatic resources for tourism in North America. Clim. Res. 27, 105–117. Shorrocks, B., 2007. The Biology of African Savannahs. Oxford University Press, Oxford. Smit, I.P.J., Grant, C.C., Devereux, B.J., 2007. Do artificial waterholes influence the way herbivores use the landscape? Herbivore distribution patterns around rivers and artificial surface water sources in a large African savanna park. Biol. Conserv. 136, 85–99. Thomas, L., et al., 2009. Distance 6.0, Release 2. Research Unit for Wildlife Population Assessment. University of St. Andrews, UK. Turpie, J., Joubert, A., 2001. Estimating potential impacts of a change in river quality on the tourism value of Kruger National Park: an application of travel cost, contingent and conjoint valuation methods. Water SA 27, 387–398. Valeix, M., Loveridge, A.J., Davidson, Z., Madzikanda, H., Fritz, H., Macdonald, D.W., 2010. How key habitat features influence large terrestrial carnivore movements: waterholes and African lions in a semi-arid savanna of north-western Zimbabwe. Landsc. Ecol. 25, 337–351. Zizka, A., Govender, N., Higgins, S.I., 2014. How to tell a shrub from a tree: a life-history perspective from a South African savanna. Austral Ecol. http://dx.doi.org/10.1111/ aec.12142.