Skip to main content
PMC home page
Ambio logoLink to Ambio
. 2018 Feb 12;47(7):781–793. doi: 10.1007/s13280-018-1020-4

Using crowd-sourced photos to assess seasonal patterns of visitor use in mountain-protected areas

Chelsey Walden-Schreiner 1,✉, Sebastian Dario Rossi 2, Agustina Barros 3, Catherine Pickering 4, Yu-Fai Leung 5,6
PMCID: PMC6188969  PMID: 29435731

Abstract

Managing protected areas effectively requires information about patterns of visitor use, but these data are often limited. We explore how geotagged photos on Flickr, a popular photo-sharing social-media site, can generate hotspot maps and distribution models of temporal and spatial patterns of use in two mountain-protected areas of high conservation value. In Aconcagua Provincial Park (Argentina), two routes to the summit of Aconcagua were used in summer, but most visitors stayed close to the main road, using formal and informal walking trails and the Visitor Centre, while in winter, there was very limited visitation. In Kosciuszko National Park (Australia), alpine walking trails were popular in summer, but in winter, most visitors stayed in the lower altitude ski resorts and ski trails. Results demonstrate the usefulness of social-media data alone as well as a complement for visitor monitoring, providing spatial and temporal information for site-specific and park-level management of visitors and potential impacts in conservation areas.

Keywords: Aconcagua Provincial Park, ALPINE, Geotagged photos, Informal trails, Kosciuszko National Park, MaxEnt modelling

Introduction

Protected areas are key mechanisms for conserving biodiversity while also providing opportunities for human–environment interactions through recreation and tourism (Newsome et al. 2012; Watson et al. 2014). These interactions offer a host of health and well-being benefits (Romagosa et al. 2015), along with economic benefits to local communities (Worboys et al. 2015). In mountain-protected areas, including parks established to conserve outstanding natural features, tourists can engage in a range of activities including mountaineering, hiking, sightseeing, skiing, bird watching, and fishing (Martínez Pastur et al. 2016; Musa et al. 2016). As a result, some mountain-protected areas are popular tourism destinations, with estimates of 15–20% of the over one billion international tourists each year visiting these regions (Debarbieux et al. 2014). Visitor use of protected areas can also have negative impacts on natural resources, including wildlife, soil, water, and vegetation (Hammitt et al. 2015). Alpine environments are particularly sensitive to impact, including from visitor-created informal trails, because of the limited ability of alpine vegetation to recover from trampling due to short growing seasons and soil characteristics (Leung et al. 2011; Barros et al. 2013). Monitoring visitor numbers and the spatial and temporal patterns of use inform management strategies intended to mitigate and minimize these negative impacts while maintaining opportunities for quality visitor experiences (Hadwen et al. 2007; Newsome et al. 2012).

Data on visitor use of protected areas can be collected through a variety of manual and digital methods, including direct observation, trail counters, and on-ground or web-based-tracking technology (Newsome et al. 2012; Eagles 2014; Shoval and Ahas 2016). However, the data collected are often limited by multiple constraints, including a shortage of personnel and/or funding to obtain detailed data across large regions, inconsistencies in how monitoring is conducted, and practical limitations associated with some technology resulting in inaccurate data (Newsome et al. 2012; Eagles 2014). A dearth of reliable visitor data for protected areas is often credited as a major limitation for proactive management (Hadwen et al. 2007; Worboys et al. 2015).

Crowd-sourced data are increasingly used as an efficient and complementary approach for visitor monitoring in protected areas (Levin et al. 2017). Crowd-sourced data are information generated by many individuals, often accessed through web-based platforms (See et al. 2016). This includes geotagged photos shared publicly on social-media and other social-media-derived sources of volunteered geographic information (i.e., GPS tracks), which can be used for tourism (Bizarro et al. 2016; Campelo and Nogueira Mendes 2016; Santos et al. 2016; Levin et al. 2017) and visitor infrastructure research (Hennig 2017; Walden-Schreiner et al. 2018). Social media provide outlets for visitors desiring to share information about their travel, including intense experiences in areas known for their conservation value, remoteness, and iconic features. Summiting peaks, especially the highest in a region, is often seen as a personal achievement (Musa et al. 2016) and sharing experiences has been found to be a key motivation in sharing photos on social media (Oeldorf-Hirsch and Shyam Sundar 2016).

Geotagged images can serve as proxies of visitor numbers, as well as reflect spatial and temporal variations in visitor use. Studies using geotagged photos have found correlations between photo numbers and empirical visit counts at large spatial scales (Wood et al. 2013; Levin et al. 2015; Spalding et al. 2017), contributed to models of economic impacts (Sonter et al. 2016), explored the distribution of cultural ecosystem services (Richards and Friess 2015; Martínez Pastur et al. 2016), and modelled visitor flows at more local levels (Orsi and Geneletti 2013). Data from popular photo-sharing sites such as Instagram, Panoramio, and Flickr are accessible through application-programming interfaces. A recent study by van Zanten et al. (2016) highlighted how Panoramio, Flickr, and Instagram revealed high spatial agreement for geotagged photos across Europe, including for mountainous terrain. Specific to protected areas, recent estimates suggest that 11% of the 20.6 million geotagged photos shared on Flickr are taken within protected areas, of which 6.4% are from remote locations (Levin et al. 2015). These areas include parks of high conservation value, such as those conserving the highest mountains on every continent. Research also suggests that social-media data provide reliable alternatives for understanding tourists’ preferences for nature-based experiences, specifically for biodiversity, with Flickr aligning best with preferences for less charismatic biodiversity (Hausmann et al. 2017).

Often, research has examined visitor counts and patterns at a large spatial scale, such as for large parks or even at national and international scales (Wood et al. 2013; Levin et al. 2015). Fewer studies, however, have examined visitor use patterns at smaller scales (Richards and Friess 2015) or examined seasonal visitor–environment interactions (Sessions et al. 2016) based on infrastructure and environmental factors, critical for understanding such interactions and mitigating impacts. The aims of this study are to assess the utility of using social-media geotagged photos to determine spatial and temporal visitation patterns in remote mountain-protected areas. Specifically, this paper: (1) examines metadata characteristics of geotagged photos; (2) compares geotagged photo locations with visitor data to assess if the photos can be used as proxies for visitation patterns in the parks; (3) analyses the temporal and spatial distributions of visitors based on geotagged photos; and (4) assesses what factors are associated with the spatial distribution of photos in the parks. The research on geotagged photos was conducted in two mountain parks of high conservation value containing iconic summits and popular for a range of tourism activities, but that differ in the types of activities available and the extent and types of infrastructure for visitors: Aconcagua Provincial Park in Argentina and Kosciuszko National Park in Australia.

Materials and methods

Study areas

Aconcagua Provincial Park in the Andes Mountains of Argentina is an IUCN Category II protected area that conserves alpine ecosystems, endemic flora and fauna, glaciers, rivers, as well as contains the highest summit in the Southern Hemisphere, Mt Aconcagua (6962 m a.s.l.) (Fig. 1a). Over 120 plant species are conserved in Aconcagua, present below 4400 m a.s.l., with most occurring in the valley floors, where campsites, infrastructure, and formal and informal trails are also located (Barros et al. 2013) (Fig. 2a, b). Only one road accesses the park, the international highway connecting Argentina and Chile, with people entering via the two valleys intersecting the road: Horcones Valley and Vacas Valley (Fig. 1a). During the summer months, approximately 40 000 people visit the park, of which around 35 000 are day visitors (Barros et al. 2015). Day visitors only use the Horcones Valley, where there is a Visitor Centre and self-guided circuit walk of 2 km with views of Mt Aconcagua (Barros et al. 2015). Mountaineers and trekkers travel further into the park to attempt the summit or reach base camps through Horcones Valley (30 km) or Vacas Valley (42 km). In winter, nearly, all visitation consists of day trippers using the Visitor Centre and an adjacent short walk (300 m) in Horcones Valley, with nearly no attempts on the summit due to severe weather conditions.

Fig. 1.

Fig. 1

Open in a new tab

Locations of the two parks in Argentina and Australia with road networks, formal and informal trails and main visitation areas. Aconcagua Provincial Park (a) is located in the Central Andes in Mendoza, Argentina and Kosciuszko National Park (b) is located in the Snowy Mountains in New South Wales, Australia

Fig. 2.

Fig. 2

Open in a new tab

Mountain-protected areas of high conservation value and popular with tourists in Argentina and Australia. a Park visitors congregate in Horcones Valley at the entrance to Aconcagua Provincial Park, Argentina. b Visitors hiking through Vacas Valley in Aconcagua Provincial Park. c Visitors hiking on a steel mesh track in Kosciuszko National Park, Australia. d View of glacial lakes in Kosciuszko National Park

In contrast, Kosciuszko National Park in southeastern Australia is lower in elevation, receives more visits, is easier to access, and is popular with visitors in winter as well as summer. The park protects unique montane and alpine habitats with several endemic and endangered species, and includes the highest summit on the continent (Mt Kosciuszko, 2228 m a.s.l.) (Fig. 1b). Kosciuszko is an IUCN Category II protected area, UNESCO Biosphere Reserve, and a biodiversity centre identified by the World Conservation Monitoring Centre with over 1.4 million visits a year (Fig. 2c, d) (Roy Morgan Research 2015). Although winter is the most popular time to visit the park for snow-based activities, including skiing in the major resorts located in the subalpine zone, walking in the alpine areas around Mt Kosciuszko is popular in summer with over 100 000 visitors in this part of the park in 2005 (Johnston and Growcock 2005). There are eight entry points, including a series of paved and unpaved roads, some of which traverse the park facilitating year round access (Fig. 2b). The summit of Mount Kosciuszko is also easily accessed by a 5–8 h walk from one of the two ski resorts when the alpine area is snow free, and by skis in winter (Pickering and Buckley 2003).

Existing visitor monitoring data

Both parks collect visitor count data, but by different methods. For Aconcagua, park rangers collect data from November to March when conditions are warmer and most people use the park. This includes data from entry permits (fees) and a registration book at the main entrance. Data include information on the number of days visitors intend to stay in the park, locations, and activities undertaken (i.e., day visit, short or long trek, climb) (Barros et al. 2015). Data from permit fees indicate that there are over 40 000 day visits to the Horcones Valley per year, but only around 5000 people travel further into the park to go hiking or climbing (Massarelli 2016). The timeframe informed and matched the date range of photos requested from Flickr.

In Kosciuszko, visitor data are collected from park entry gates, surveys, and traffic counters on several roads and trails (Worboys and Pickering 2002), at different temporal and spatial scales. Data are often fragmentary due to resource limitations and occasional equipment malfunctions. Therefore, limited visitor data were available for this analysis for the park.

Crowd-sourced visitor data

Crowd-sourced geotagged photos from the two parks were obtained from Flickr. This source of geotagged photos has been used for other research, including in protected areas, and has an accessible application-programming interface (Alivand and Hochmair 2017). All publicly shared photos on Flickr containing a geotag over several years and seasons of use (1 November 2010 and 31 March 2016) were requested from the Flickr API on 14 September 2016, and located within the boundaries of the parks, were obtained using the application-programming interface. Photo metadata included where and when it was taken (longitude and latitude), when it was uploaded, owner ID, image URL, and camera type. Data were written to comma-separated values (csv) files for statistical analysis and also converted to geographic information system (GIS) as shapefiles.

Two of the authors with extensive site knowledge of both parks examined the photos to determine whether they were correctly geotagged. Specifically, all photos more than 250 m from park features (i.e., roads, trails, visitor centres, and ski resorts) were inspected. If an image was no longer accessible (i.e., owner had removed or changed viewing permissions after the metadata were captured in the sample), the title and text tags were checked. Only 3.3% (33 photos) for Aconcagua and 0.17% (10 photos) for Kosciuszko appeared to be in the wrong location, including outside the park, or not taken from the ground (i.e., from a plane). For Aconcagua, the content of the photos was diverse including landscapes, natural features, campsites, and people hiking or sightseeing. The potential outliers were clearly incorrect (e.g., showing the international highway outside of the park) and, therefore, removed from the data set. The potential outliers from Kosciuszko were retained as it is possible, and common, for visitors bushwalking or participating in snow-based activities to be far from formal infrastructure and, therefore, difficult to determine if they were actually outside the large park boundaries or if the photos were in the wrong location. Ten MaxEnt iterations were performed, each using a different random sample of photo data, to converge on a distribution model solution. This was also an attempt to mitigate the influence of potentially incorrect geotags.

Analyses

Visitation statistics and correlations

Visitor statistics recorded on ground from the two parks were compared with the number of geotagged Flickr photos to assess if the photos reflected empirical visitation patterns. For Aconcagua, Spearman’s rho correlations compared data for the 5 month visitor season over the 5 year period, where visitor count from park registries and permits were available (n.b., aggregated by week for analysis). Spearman’s rho correlations were used as data violated the assumptions of parametric tests. For Kosciuszko, data from infrared trail counters and park ranger observations were descriptively compared to Flickr photos as visitor data were only available to the authors for specific days (i.e., Easter Sunday) or weeks (i.e., time period between the Christmas and New Year holidays).

Visitor spatial distribution

To assess patterns of visitor use, including use of park infrastructure and environmental features, GIS data were obtained from the park management agencies. Data included visitor infrastructure (e.g., roads, trails, informal trails (Aconcagua only), campsites, buildings, and parking areas) and environmental characteristics [e.g., vegetation cover types, elevation, slope, rivers, and location of glaciers (Aconcagua only)]. For Kosciuszko, the boundaries of the ski resorts were also included, as well as other facilities (e.g., toilets, changing rooms, and offices), which were part of the buildings layer for the park.

To analyse the spatial and temporal distributions of visitors, and visitors’ use of infrastructure and environmental features, a near distance analyses was conducted in ArcGIS 10.3 for the two parks separately. This was done for all data, and separately for summer (December through February) and winter (June through August). Features included infrastructure (i.e., bridges, buildings, and facility points), campsites, formal and informal trails (Aconcagua only), boundaries of two ski resorts (Kosciuszko only), and other environmental features accessible to visitors (i.e., rivers for Kosciuszko and glaciers for Aconcagua) and salient to managers due to environmental impact concerns (Walden-Schreiner et al. 2018). Trail layers in Aconcagua included formal trails designated by the park agency and informal trails created by users and mapped in the field using a Garmin Oregon 450 GPS (Barros and Pickering 2017). For Kosciuszko, trails were classified as either hardened, unhardened, or designated for snow activities (i.e., cross-country skiing, snowshoeing) based on metadata or confirmed by the park’s website.

The machine-learning algorithm MaxEnt further examined the associations between multiple environmental and infrastructure factors and the spatial distribution of visitors assumed from geotagged photos. MaxEnt is a distribution modelling approach that examines covariate conditions at presence locations (i.e., geotagged photo location) and compares it to random locations in the greater area of potential occurrence (i.e., park boundaries) (Elith et al. 2011). Studies demonstrate MaxEnt to be successful in modelling species distributions of flora (Koch et al. 2017) and fauna (Morán-Ordóñez et al. 2017), including with limited sample sizes (Elith et al. 2011). The algorithm assumes data coverage is incomplete and provides probability of presence based on combinations of factors for locations in the broader area of potential occurrence. Model outputs include maps of the probability of occurrence, ranking of variable importance, variable response curves, and evaluations of the models fit.

It is important to note that MaxEnt model results, specifically response curve interpretation, may be complicated by strongly correlated variables if many variables changing together elicit change in another variable. To address this, individual MaxEnt models created with only one variable were also examined to explore variable values and predicted presence. Halvorsen et al. (2016) point out that spatial dependence among predictor variables can be part of the overall landscape and results from their study suggest that MaxEnt models were unaffected in terms of predictive performance.

MaxEnt has explored distributions of visitors (Westcott and Andrew 2015), recreation impacts (Braunisch et al. 2011; Coppes and Braunisch 2013), and types of cultural ecosystem services (Richards and Friess 2015) within protected areas. As with any modelling approach, MaxEnt is sensitive to model parameters, specifically the selection of background data for the potential area of occurrence. To address these concerns, this analysis followed recommendations in the literature as summarized by Walden-Schreiner et al. (2017). For vector data, the factor was characterized by the distance to the polygon (e.g., campsite and ski resort) or line (e.g., trail and road) feature for the model. For continuous data (e.g., elevation), the model used the value at that location. The overall contribution of each factor was evaluated through permutation importance, which reflects the change to model fit when values of a specific factor are randomly permuted. The resulting percentages do not depend on the path taken by the algorithm to arrive at a solution. MaxEnt model fit is described by the area under the curve (AUC) for the receiver operator characteristics (ROC), with values above 0.75 considered acceptable and a maximum value of 1 indicating perfect prediction of presence (Richards and Friess 2015).

Results

Flickr data characteristics

A total of 981 photos for Aconcagua and 5965 photos for Kosciuszko were obtained in September 2016 from Flickr for the specified sample period. Clear seasonal patterns emerged based on when the photos were taken for both parks. Nearly, 80% of the Aconcagua photos were taken between December and March during the warmest months, while both the ski season (June to September) and summer (December to end of February) were both popular times for photos for Kosciuszko, along with major public holidays such as Easter in autumn (March/April) (Fig. 3). Photos from Aconcagua were uploaded by 130 unique user accounts, with a median of 2 photos in the park per users. In Kosciuszko, photos were posted from 316 unique user accounts with a median of 3 photos per user in the park.

Fig. 3.

Fig. 3

Open in a new tab

Total number of photos taken per month for Aconcagua Provincial Park and Kosciuszko National Park from November 2010 to March 2016

Comparisons with visitor counts

During the sampling period, there were 199 387 visits to Aconcagua, 27 706 of which involved hiking between one and 20 days and 171 681 used the Visitor Centre and the short walk in the Horcones Valley. When visitation data were compared to Flickr photo data using Spearman’s rho, there were significant correlations between the number of photos and the total number of visits to Aconcagua (r = 0.249, p < 0.05), as well as the number of photos and the number of hikers (r = 0.312, p < 0.05).

Data from trail counters in the alpine area around the summit of Kosciuszko estimated over 4400 people visited the summit of Mt. Kosciuszko during the popular 5-day period in summer between Christmas and New Year’s (i.e., 25 December 2015 to 1 January 2016). Other popular periods include Easter Sunday when over 1500 visitors were observed hiking the summit and more than 700 visitors hiking in the alpine area on 27 March 2016 alone (Christopher Darlington, pers. com. 2016). For the ski resorts, winter is popular, including late in the season when there are major school holidays in August. The number of Flickr photos followed a similar trend, with clear increases associated with holidays in December/January (i.e., Christmas, New Year’s), March/April (Easter) within the alpine area, and in the resorts in August when there is snow on the ground and coinciding with 2 weeks of school holidays (Fig. 3).

Visitor distributions

Based on spatial coordinates extracted from geotagged photos, most visitors to the parks were on or near visitor infrastructure (Fig. 4a) (n.b., the following mentions of photos refer to the spatial locations). Kernel density maps of Aconcagua in summer revealed hotspots of use near the Visitor Centre, as well as the main trails and campsites on the way to the summit of Mt. Aconcagua. Overall, most photo locations in Aconcagua were within 5 m of a feature (62%), with the closest features formal trails (30%), the Visitor Centre (22%), and campsites (17%), and some use of informal trails (14%) (Fig. 4b). During the summer months, the closest features were formal trails (38%) and campsites (24%), while in winter, it was the Visitor Centre (67%).

Fig. 4.

Fig. 4

Open in a new tab

Number of Flickr photos per distance band for the two parks (a) and frequency of the closest park feature for Aconcagua Provincial Park (APP) (b) and Kosciuszko National Park (KNP) (c) for the summer (December through February) and winter (June through August) months

Over half of all photos from Kosciuszko were also taken within 5 m of a feature (54%), with the closest features the ski resorts (49%) and major roads (13%) (Fig. 4a). There were also clear differences in the distribution of photos between summer and winter in the park (Fig. 4c). In summer, the ski resorts that are open year round were popular (30%), with people golfing, walking, sightseeing, and going to cafes. However, visitors used other features in summer such as hardened walking trails (22%) often in the alpine area, or went to the rivers which are popular for fishing and camping (16%). In winter, nearly three-quarters of all photos were taken closest to ski resorts (74%), followed by rivers (10%) and roads (6%).

Factors associated with visitor distributions

When modelled using MaxEnt to explore combinations of factors in predicting presence beyond sampled locations (i.e., locations, where photos were taken), similar patterns emerged to those found with the descriptive summaries of the closets features to photos (Fig. 5). For Aconcagua in summer, models indicated high probabilities that photos were taken near the Visitor Center and the day use area in the Horcones Valley at the southern boundary of the park. In addition, high probabilities extended into the park along the two trails used to access the Mt Aconcagua summit (Fig. 5a, b). Clusters of high probabilities of presence in winter were almost entirely restricted to the Visitor Centre in Horcones Valley, as the park is effectively closed to hiking and climbing, with the exception of short trails near the Visitor Centre (Fig. 5c, d).

Fig. 5.

Fig. 5

Open in a new tab

Presence of photos based on logistic probability modelling for Aconcagua Provincial Park (left) and Kosciuszko National Park (right) during the summer (December through February) and winter (June through August) months. Warmer colors (yellow to red) indicate greater predicted probabilities of photo presence

During the summer in Kosciuszko, the popularity of hiking trails from the road leading to the summit of Mt Kosciuszko and other hikes in the subalpine and alpine area in the southern third of the park (co-located in and near ski resorts) are apparent as hotspot lines, with clusters in other key destinations such as Yarrangobilly Caves in the middle of the park (Fig. 5e, f). In winter, probability hotspots are more concentrated in and around the ski resorts (Fig. 5g, h), with limited use of areas further away in the southern part of the park, likely by visitors snow camping, cross-country skiing, and/or snowboarding.

For both parks, there were differences in the most important explanatory variables from the MaxEnt modelling, reflecting key differences in infrastructure use patterns and locations between winter and summer, as well as between the two parks (Table 1). In Aconcagua, formal trails were the most important explanatory factor in summer (permutation importance of 80.6%), with buildings (i.e., Visitor Centre) the second most important factor (9.2%). Informal trails contributed little to the summer model (3.3%). In winter, a different pattern was observed with buildings (i.e., Visitor Centre) the most important explanatory factor (64%), and informal trails more important (11.5%) than formal trails (0.2%), reflecting more dispersed use of the flat open areas around the entrance to the park. With respect to environmental factors, slope contributed over 30% to the autumn and spring models, with increased probabilities of presence at slope values of less than five degrees. This finding coinciding with terrain conditions near the Visitor Centre.

Table 1.

Permutation importance of factors included in the MaxEnt models for Aconcagua Provincial Park and Kosciuszko National Park during the summer and winter months

Aconcagua Provincial Park Kosciuszko National Park
Summer Autumn Winter Spring All Summer Autumn Winter Spring All
Dec–Feb Mar–May June–Aug Sep–Nov Dec–Feb Mar–May June–Aug Sep–Nov
Permutation importance Permutation importance
Infrastructure Infrastructure
Formal trail 80.6 3.1 0.2 0.6 63.7 Hardened trail 26.7 18.7 3.8 5.5 3.5
Unhardened trail 10.9 3.2 4.6 17.2 9.6
Informal trail 3.3 27.1 11.5 10.6 7.8 Informal trail Not applicable
Ski trails Not applicable Ski trails 3.9 22.1 22.2 9.1 9.4
Road 1.1 1 9.1 0.3 3 Road 16.8 33.1 21.6 7.2 22.4
Parking 0.6 0.2 0 0.1 0.4 Parking 7.8 2.7 6.6 19.1 7.8
Ski resorts Not applicable Ski resorts 14.8 13.2 21.1 22.9 32.3
Building 9.2 27.3 63.6 44.2 15.7 Building 5.2 1.6 2 5.7 4.8
Campsite 1.8 0.2 0.3 0.1 3 Campsite 2 0.7 2.4 1 1.9
Environmental Environmental
Elevation 0.4 1 0 6 0.6 Elevation 4.9 1.1 9 3.1 5.2
Slope 2.8 34.9 15.1 37.4 5.8 Slope 3.1 1.7 0.8 3.1 0.7
Vegetation type 0 0 0.1 0.4 0 Vegetation type 2.2 0.5 2.1 1.4 1.2
River 0.2 0.5 0 0 0.1 River 1.8 1.3 3.8 4.8 1.2
Glacier 0 4.7 0 0.2 0 Glacier Not applicable
AUC 0.98 0.991 0.96 0.996 0.978 AUC 0.947 0.962 0.969 0.959 0.936
(St. Dev.) (0.002) (0.006) (0.055) (0.002) (0.003) (St. Dev.) (0.009) (0.012) (0.006) (0.010) (0.007)
Open in a new tab

In Kosciuszko, there were also differences between summer and winter in the importance of various factors. In summer, hardened (26.7%) and unhardened trails (10.9%) ranked among the most important explanatory variables, along with roads (16.8%) and ski resorts (14.8%). In winter, ski trails (22.2%), roads (21.6%), and ski resorts (21.1%) contributed collectively to almost two-thirds of the model, with hardened and unhardened trails far less important, reflecting differences in recreational opportunities and access in the park between periods with, and without, snow cover. Environmental factors contributed 5% or less across all models for Kosciuszko. In the overall model, elevation was the highest ranked environmental factor at 5.2%, with greater probabilities of presence predicted at elevations in and around ski resorts.

Discussion

Geotagged social-media photos can provide protected area managers with timely and useful data about spatial–temporal patterns of use and the popularity of different types of infrastructure as a complement to, and in concert with, other visitor monitoring approaches. The availability (Levin et al. 2015), and relative spatial accuracy (Zielstra and Hochmair 2013) of social-media data, is particularly important in situations when resources and/or remote locations limit on ground visitor monitoring. Therefore, access to social-media data on how visitors access, use, and value, different areas and infrastructure in protected areas will be increasingly critical for those responsible for their management (Hausmann et al. 2017; Heikinheimo et al. 2017).

Flickr data can provide a proxy for visitation counts at the park scale as documented in other studies (Wood et al. 2013; Levin et al. 2015, 2017; Sonter et al. 2016), as well as provide detailed site and infrastructure-specific information of where people go and when in large and remote mountainous protected areas. The significant correlations found between the number of photos and the visitation numbers in this study offer further evidence of support, even though the correlation values are lower than those reported in Wood et al. (2013) and Keeler et al. (2015), possibly due to smaller sample sizes in larger, more remote areas in our study.

The results for Aconcagua and Kosciuszko highlight how Flickr data can provide insights into spatial and seasonal patterns of visitor use. With nearly, 97% correctly geotagged to the parks, Flickr photos provide increasingly accurate data on visitor use of these areas. With the increasing use of mobile devices such as smartphones globally (Poushter 2016), social-media data provided by visitors are only likely to further increase in quality, accuracy, volume, and area. This can be seen already for the two parks as over the more than 5 years of data, there was increasing use of mobile devices to capture images shared on Flickr, potentially reducing any inaccuracies introduced when manually geotagging digital photographs.

The spatial and temporal patterns of use for the parks differed reflecting differences in access, recreational opportunities, and facilities between the parks and over seasons within each park. Aconcagua is a summer use park, mainly by short stay visitors using walking trails and the Visitor Centre, with the use of the rest of the park limited to the two main routes to the summit. In contrast, far more of Kosciusko is used year round by visitors, but with differences in where people go and which facilities they use between winter and summer as snow cover influences what visitors do and where.

The MaxEnt distribution models provide more detailed analysis of the relative importance of different types of infrastructure in shaping patterns of use: a critical issue for managers when deciding what types of infrastructure to provide and how to minimize resource impacts. In summer, both formal (Aconcagua) and hardened (Kosciuszko) trails ranked as the top explanatory factors in the models, reflecting how well they are used by visitors. Such concentrated use of trails designed and provided by parks agencies is heartening for managers, as trails are a major way in which they try to minimize damage from tourism (Ballantyne and Pickering 2015). These trails can enhance the visitor experience by providing comfortable and safe access to desirable sites while also minimizing impacts including from trampling of sensitive and easily damaged environments off trails (Ballantyne and Pickering 2015; Barros and Pickering 2017). This is particularly important in alpine areas, where the recovery of vegetation from damage, including by trampling off trails, is notoriously slow and limited (Pickering and Growcock 2009).

In winter in Kosciuszko, there was concentrated visitor use of ski resorts and ski trails. Ski resorts in mountains are very popular sites for tourism, including in Australia (Morrison and Pickering 2013). However, they do have a wide range of environmental impacts which need to be minimized and managed. In addition, with reductions in snow cover in this and many other mountain regions as a result of climate change, there are new economic, social, environmental, and management issues with visitor use of mountain resorts (Morrison and Pickering 2013). As a result, some mountain resorts, including Kosciuszko as well as in North America and Europe, are increasingly promoting new activities in summer, such as mountain biking, to help offset reductions in winter use. Changes in where people go, when, and for what, create new challenges for conservation. Consequently, timely data on visitor use of these areas will be increasingly important when modelling and managing responses to changes in snow cover and the popularity of specific locations and activities. Exploring how climate conditions, like fluctuating winter snowfall, influence visitation estimates derived from social media is an important area for future research.

A limitation inherent to geotagged photos and other social-media-based data is sampling bias (Richards and Friess 2015; Levin et al. 2017). Not all visitors take and share photos, and for those that do, there are a range of photo-sharing platforms available that vary in their popularity, aims and if the data are publicly available. In addition, the availability and use of smartphones and other devices to take and share photos online vary among and within countries (Poushter 2016). As a result, geotagged photos and other social-media data will not be representative of all visitors, all activities, or all types of people engaged in a given activity in protected areas. Some sites and activities are also more ‘shareable’ than others, resulting in additional biases. Photos from the two parks assessed here, for example, were more likely to be of stunning natural landscapes than restrooms or carparks. This study attempted to minimize the impacts of this phenomenon by leveraging MaxEnt, which assumes that data coverage is incomplete and provides predicted probabilities of presence based on combinations of factors. However, it may still miss site-specific patterns in some locations and activities. Indeed, future research efforts can explore the photo content of shared pictures providing useful insights for monitoring; assessing cultural ecosystem services and potentially identifying specific visitor impact concerns such as wildlife–visitor interactions.

Spatial accuracy can be a concern when leveraging volunteered geographic information, especially geotagged photos where users may manually assign the photo’s location. Previous research has found visitors have tagged the photo’s contents as the location as opposed to where it was taken from (Orsi and Geneletti 2013). The analysis of the photos in this study revealed that less than 3% were inaccurately geotagged, yet the accuracy could still be within several meters depending on the device used. For formal and informal trails very closely collocated, challenges in interpreting where a visitor was standing when taking a photo could arise without examination of the photo contents. However, for many of the informal trails in this study, the marked separation from the formal trails reduced this concern and it is important to capture considering the fragmentation effects associated with this type of visitor impact and the spatial accuracy identified in other studies (Antoniou et al. 2010).

Geotagged photos provide point or area of interest information, and, therefore, do not capture the route a visitor uses. Research on other platforms that share GPS route data such as Strava, GPSies, Wikiloc, and MapMyRide are also increasingly being used to study visitor use of protected areas. This includes generating hotspot maps of trail use by visitors engaged in different activities and use of informal trail networks (Campelo and Nogueira Mendes 2016; Santos et al. 2016; Korpilo et al. 2017). Finally, some studies are using a combination of onsite visitor surveys and social-media data or public partition GIS (PPGIS) to assess how different visitor groups use trails, including for visitor conflict (Wolf et al. 2017) and the influence of natural features and visitor infrastructure quality on visitor experience (Pietilä and Kangas 2015).

Conclusion

This study highlighted how geotagged social-media data, specifically photographs from Flickr, can support and complement on ground visitor use and impact monitoring in mountain-protected areas of high conservation value. Social-media data contribute insights into seasonal and temporal patterns of visitor use. This includes the ability to detect spatial use patterns of informal trails in Aconcagua and dispersed winter use in Kosciuszko, which otherwise would be difficult or not possible to capture through park registration permits or trail counters. The availability of social-media data, even in light of the limitations discussed, provides opportunities to address urgent and dynamic management challenges associated with visitor use in protected areas when data may otherwise be unavailable or limited. Understanding patterns of visitor use and how infrastructure and environmental conditions influence such patterns helps inform monitoring programs and management strategies to protect high conservation value resources and provide for quality visitor experiences in these and other popular parks.

Acknowledgements

We thank Ruben Massarelli from Aconcagua Provincial Park who provided us with visitor statistics for the park and Dirección de Recursos Naturales Renovables, Mendoza, for providing the GIS layers for Aconcagua. We also thank Tim Greenville, Pamela O’Brien, and Anthony Evans for providing traffic counter location data and information regarding visitor use patterns in Kosciuszko National Park. Thank you also to Will Dorrington for providing GIS layers for Kosciuszko National Park, and Christopher Darlington for providing data on visitor use of the alpine area in the park.

Biographies

Chelsey Walden-Schreiner

is a research associate at North Carolina State University, USA. Her research interests include recreation ecology, visitor use and impact monitoring, and applications of geospatial technologies for conservation and protected area management.

Sebastian Dario Rossi

is a researcher in Human Geography at the Desertification and Land Management Laboratory (LaDyOT) at CONICET, Argentina. His research interests include protected areas planning and management, outdoor recreation, and nature–society interactions in natural areas including environmental psychology and place making.

Agustina Barros

is a researcher in Ecology at the Instituto Argentino de Nivología y Glaciología y Ciencias Ambientales (IANIGLA), Centro Científico Tecnológico (CCT) CONICET, Mendoza, Argentina. Her research interests include recreation and alpine ecology, tourism, and protected area management.

Catherine Pickering

is a professor and researcher in the School of Environment, Griffith University, Australia. Her research interests include ecology, tourism, and protected area management.

Yu-Fai Leung

is a professor at North Carolina State University, USA. His research interests include sustainable tourism, monitoring, geospatial analysis, and protected area management.

Contributor Information

Chelsey Walden-Schreiner, Email: cwalden1650@gmail.com.

Sebastian Dario Rossi, Phone: +54 261 524-4113, Email: sebastian.rossi@griffithuni.edu.au, Email: srossi@mendoza-conicet.gob.ar.

Agustina Barros, Phone: +54 261 5244235, Email: anaagustinabarros@gmail.com, Email: abarros@mendoza-conicet.gob.ar.

Catherine Pickering, Email: c.pickering@griffith.edu.au.

Yu-Fai Leung, Email: Leung@ncsu.edu.

References

  1. Alivand M, Hochmair HH. Spatiotemporal analysis of photo contribution patterns to Panoramio and Flickr. Cartography and Geographic Information Science. 2017;44:170–184. doi: 10.1080/15230406.2016.1211489. [DOI] [Google Scholar]
  2. Antoniou V, Morley J, Haklay M. Web 2.0 geotagged photos: Assessing the spatial dimension of the phenomenon. Geomatica. 2010;64:99–110. [Google Scholar]
  3. Ballantyne M, Pickering C. The impact of trail infrastructure on vegetation and soils: Current literature and future directions. Journal of Environmental Management. 2015;164:53–64. doi: 10.1016/j.jenvman.2015.08.032. [DOI] [PubMed] [Google Scholar]
  4. Barros A, Pickering C. How networks of informal trails cause landscape level damage to vegetation. Environmental Management. 2017;60:57–68. doi: 10.1007/s00267-017-0865-9. [DOI] [PubMed] [Google Scholar]
  5. Barros A, Gonnet J, Pickering C. Impacts of informal trails on vegetation and soils in the highest protected area in the Southern Hemisphere. Journal of Environmental Management. 2013;127:50–60. doi: 10.1016/j.jenvman.2013.04.030. [DOI] [PubMed] [Google Scholar]
  6. Barros A, Pickering C, Gudes O. Desktop analysis of potential impacts of visitor use: A case study for the highest park in the Southern Hemisphere. Journal of Environmental Management. 2015;150:179–195. doi: 10.1016/j.jenvman.2014.11.004. [DOI] [PubMed] [Google Scholar]
  7. Bizarro, S., R.M. Nogueira Mendes, T. Santos, and R.P. Julião. 2016. Monitoring outdoor recreation in Serra da Estrela Natural Park, Portugal. In D. Vasiljević et al., eds. Proceedings of the 8th International Conference on Monitoring and Management of Visitors in Recreational and Protected Areas: (Vol. MMV 8—Proceedings). 8th International Conference on Monitoring and Management of Visitors in Recreational and Protected Areas. Novi Sad, Serbia. 26–30 September. pp. 137–139.
  8. Braunisch V, Patthey P, Arlettaz R. Spatially explicit modeling of conflict zones between wildlife and snow sports: Prioritizing areas for winter refuges. Ecological Applications. 2011;21:955–967. doi: 10.1890/09-2167.1. [DOI] [PubMed] [Google Scholar]
  9. Campelo MB, Nogueira Mendes RM. Comparing webshare services to assess mountain bike use in protected areas. Journal of Outdoor Recreation and Tourism. 2016;15:82–88. doi: 10.1016/j.jort.2016.08.001. [DOI] [Google Scholar]
  10. Coppes J, Braunisch V. Managing visitors in nature areas: Where do they leave the trails? A spatial model. Wildlife Biology. 2013;19:1–11. doi: 10.2981/12-054. [DOI] [Google Scholar]
  11. Debarbieux, B., M. Oiry Varacca, G. Rudaz, D. Maselli, T. Kohler, and M. Jurek (eds.). 2014. Tourism in mountain regions: Hopes, fears and realities. Sustainable Mountain Development Series. Geneva, Switzerland: UNIGE, CDE, SDC.
  12. Eagles PFJ. Research priorities in park tourism. Journal of Sustainable Tourism. 2014;22:528–549. doi: 10.1080/09669582.2013.785554. [DOI] [Google Scholar]
  13. Elith Jane, Phillips Steven J., Hastie Trevor, Dudík Miroslav, Chee Yung En, Yates Colin J. A statistical explanation of MaxEnt for ecologists. Diversity and Distributions. 2010;17(1):43–57. doi: 10.1111/j.1472-4642.2010.00725.x. [DOI] [Google Scholar]
  14. Hadwen WL, Hill W, Pickering C. Icons under threat: Why monitoring visitors and their ecological impacts in protected areas matters. Ecological Management and Restoration. 2007;8:177–181. doi: 10.1111/j.1442-8903.2007.00364.x. [DOI] [Google Scholar]
  15. Halvorsen R, Mazzoni S, Dirksen JW, Næsset E, Gobakken T, Ohlson M. How important are choice of model selection method and spatial autocorrelation of presence data for distribution modelling by MaxEnt? Ecological Modelling. 2016;32:108–118. doi: 10.1016/j.ecolmodel.2016.02.021. [DOI] [Google Scholar]
  16. Hammitt WE, Cole DN, Monz C. Wildland recreation: Ecology and management. 3. Chichester, UK: Wiley Blackwell; 2015. [Google Scholar]
  17. Hausmann Anna, Toivonen Tuuli, Slotow Rob, Tenkanen Henrikki, Moilanen Atte, Heikinheimo Vuokko, Di Minin Enrico. Social Media Data Can Be Used to Understand Tourists’ Preferences for Nature-Based Experiences in Protected Areas. Conservation Letters. 2017;11(1):e12343. doi: 10.1111/conl.12343. [DOI] [Google Scholar]
  18. Heikinheimo V, Di Minin E, Tenkanen H, Hausmann A, Erkkonen J, Toivonen T. User-generated geographic information for visitor monitoring in a national park: A comparison of social media data and visitor survey. ISPRS International Journal of Geo-Information. 2017;6:85. doi: 10.3390/ijgi6030085. [DOI] [Google Scholar]
  19. Hennig S. OpenStreetMap used in protected area management. The example of recreational infrastructure in Berchtesgaden National Park. Eco-mont. 2017;9:30–41. [Google Scholar]
  20. Johnston SW, Growcock A. Visiting the Kosciuszko alpine area: Visitor numbers, characteristics and activities. Gold Coast, Australia: CRC; 2005. [Google Scholar]
  21. Keeler BL, Wood SA, Polasky S, Kling C, Filstrup CT, Downing JA. Recreational demand for clean water: Evidence from geotagged photographs by visitors to lakes. Frontiers in Ecology and the Environment. 2015;13:76–81. doi: 10.1890/140124. [DOI] [Google Scholar]
  22. Koch R, Almeida-Cortez JS, Kleinschmit B. Revealing areas of high nature conservation importance in a seasonally dry tropical forest in Brazil: Combination of modelled plant diversity hot spots and threat patterns. Journal for Nature Conservation. 2017;35:24–39. doi: 10.1016/j.jnc.2016.11.004. [DOI] [Google Scholar]
  23. Korpilo S, Virtanen T, Lehvävirtaa S. Smartphone GPS tracking—Inexpensive and efficient data collection on recreational movement. Landscape and Urban Planning. 2017;157:608–617. doi: 10.1016/j.landurbplan.2016.08.005. [DOI] [Google Scholar]
  24. Leung Y-F, Newburger T, Jones M, Kuhn B, Woiderski B. Developing a monitoring protocol for visitor-created informal trails in Yosemite National Park, USA. Environmental Management. 2011;47:93–106. doi: 10.1007/s00267-010-9581-4. [DOI] [PubMed] [Google Scholar]
  25. Levin N, Kark S, Crandall D. Where have all the people gone? Enhancing global conservation using night lights and social media. Ecological Applications. 2015;25:2153–2167. doi: 10.1890/15-0113.1. [DOI] [PubMed] [Google Scholar]
  26. Levin N, Lechner AM, Brown G. An evaluation of crowdsourced information for assessing the visitation and perceived importance of protected areas. Applied Geography. 2017;79:115–126. doi: 10.1016/j.apgeog.2016.12.009. [DOI] [Google Scholar]
  27. Martínez Pastur G, Peri PL, Lencinas MV, García-Llortente M, Martín-López B. Spatial patterns of cultural ecosystem services provision in Southern Patagonia. Landscape Ecology. 2016;31:383–399. doi: 10.1007/s10980-015-0254-9. [DOI] [Google Scholar]
  28. Massarelli. 2016. Estadisticas del Parque Provincial Aconcagua. Unpublished report. Direccion de Recursos Naturales Renovables, Gobierno de Mendoza, Argentina.
  29. Morán-Ordóñez A, Briscoe NJ, Wintle BA. Modelling species responses to extreme weather provides new insights into constraints on range and likely climate change impacts for Australian mammals. Ecography. 2017;40:1–13. doi: 10.1111/ecog.02974. [DOI] [Google Scholar]
  30. Morrison C, Pickering CM. Perceptions of the ski tourism industry and others to impacts, adaptation and limits to adaption to climate change in the Australian Alps. Journal of Sustainable Tourism. 2013;21:173–191. doi: 10.1080/09669582.2012.681789. [DOI] [Google Scholar]
  31. Musa G, Carr A, Higham J. Mountaineering tourism. Oxford, UK: Routledge; 2016. [Google Scholar]
  32. Newsome D, Moore SA, Dowling RK. Natural area tourism: Ecology, impacts and management. Bristol: Channel View Publications; 2012. [Google Scholar]
  33. Oeldorf-Hirsch A, Shyam Sundar S. Social and technological motivations for online photo sharing. Journal of Broadcasting and Electronic Media. 2016;60:624–642. doi: 10.1080/08838151.2016.1234478. [DOI] [Google Scholar]
  34. Orsi F, Geneletti D. Using geotagged photographs and GIS analysis to estimate visitor flows in natural areas. Journal for Nature Conservation. 2013;21:359–368. doi: 10.1016/j.jnc.2013.03.001. [DOI] [Google Scholar]
  35. Pickering C, Buckley R. Swarming to the summit. Mountain Research and Development. 2003;23:230–233. doi: 10.1659/0276-4741(2003)023[0230:STTS]2.0.CO;2. [DOI] [Google Scholar]
  36. Pickering C, Growcock AJ. Impacts of experimental trampling on tall alpine herbfields and subalpine grasslands in the Australian Alps. Journal of Environmental Management. 2009;91:532–540. doi: 10.1016/j.jenvman.2009.09.022. [DOI] [PubMed] [Google Scholar]
  37. Pietilä M, Kangas K. Examining the relationship between recreation settings and experiences in Oulanka National Park—A spatial approach. Journal of Outdoor Recreation and Tourism. 2015;9:26–36. doi: 10.1016/j.jort.2015.03.004. [DOI] [Google Scholar]
  38. Poushter, J. 2016. Smartphone ownership and internet usage continues to climb in emerging economies. Washington D.C.: Pew Research Center. Retrieved 18 September 2017. http://www.diapoimansi.gr/PDF/pew_research%201.pdf.
  39. Richards DR, Friess DA. A rapid indicator of cultural ecosystem service usage at a fine spatial scale: Content analysis of social media photographs. Ecological Indicators. 2015;53:187–195. doi: 10.1016/j.ecolind.2015.01.034. [DOI] [Google Scholar]
  40. Romagosa F, Eagles PFJ, Lemieux CJ. From the inside out to the outside in: Exploring the role of parks and protected areas as providers of human health and well-being. Journal of Outdoor Recreation and Tourism. 2015;10:70–77. doi: 10.1016/j.jort.2015.06.009. [DOI] [Google Scholar]
  41. Roy Morgan Research LTD. 2015. Annual visits to PWG managed parks in New South Wales. Brisbane, QLD. Retrieved 24 July 2017 from: http://www.environment.nsw.gov.au/resources/research/2014-nsw-parks-visitation-survey-report.pdf.
  42. Santos T, Nogueira Mendes RM, Vasco A. Recreational activities in urban parks: Spatial interactions among users. Journal of Outdoor Recreation and Tourism. 2016;15:1–9. doi: 10.1016/j.jort.2016.06.001. [DOI] [Google Scholar]
  43. See L, Mooney P, Foody G, Bastin L, Comber A, Estima J, Fritz S, Kerle N, et al. Crowdsourcing, citizen science or volunteered geographic information? The current state of crowdsourced geographic information. ISPRS International Journal of Geo-Information. 2016;5:55. doi: 10.3390/ijgi5050055. [DOI] [Google Scholar]
  44. Sessions C, Wood SA, Rabotyagov S, Fisher DM. Measuring recreational visitation at U.S. National Parks with crowd-sourced photographs. Journal of Environmental Management. 2016;183:703–711. doi: 10.1016/j.jenvman.2016.09.018. [DOI] [PubMed] [Google Scholar]
  45. Shoval N, Ahas R. The use of tracking technologies in tourism research: A review of the first decade. Tourism Geographies. 2016;18:587–606. doi: 10.1080/14616688.2016.1214977. [DOI] [Google Scholar]
  46. Sonter LJ, Watson KB, Wood SA, Ricketts TH. Spatial and temporal dynamics and value of nature-based recreation, estimated via social media. PLoS One. 2016;11:0162372. doi: 10.1371/journal.pone.0162372. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Spalding M, Burke L, Wood SA, Ashpole J, Hutchison J, zu Ermgassen P. Mapping the global value and distribution of coral reef tourism. Marine Policy. 2017;82:104–113. doi: 10.1016/j.marpol.2017.05.014. [DOI] [Google Scholar]
  48. van Zanten BT, van Berkel DB, Meetemeyer RK, Smith JW, Tieskens KF, Vergurg PH. Continental scale quantification of landscape values using social media data. Proceedings of the National Academy of Sciences of the United States of America. 2016;113:12974–12979. doi: 10.1073/pnas.1614158113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Walden-Schreiner C, Leung Y-F, Kuhn T, Newburger T, Tsai W-L. Environmental and managerial factors associated with pack stock distribution in high elevation meadows: Case study from Yosemite National Park. Journal of Environmental Management. 2017;193:52–63. doi: 10.1016/j.jenvman.2017.01.076. [DOI] [PubMed] [Google Scholar]
  50. Walden-Schreiner C, Leung Y-F, Tateosian L. Digital footprints: Incorporating crowdsourced geographic information for protected area management. Applied Geography. 2018;90:44–54. doi: 10.1016/j.apgeog.2017.11.004. [DOI] [Google Scholar]
  51. Watson JEM, Dudley N, Segan DB, Hockings M. The performance and potential of protected areas. Nature. 2014;515:67–73. doi: 10.1038/nature13947. [DOI] [PubMed] [Google Scholar]
  52. Westcott F, Andrew ME. Spatial and environmental patterns of off-road vehicle recreation in a semi-arid woodland. Applied Geography. 2015;62:97–106. doi: 10.1016/j.apgeog.2015.04.011. [DOI] [Google Scholar]
  53. Wolf I, Brown G, Wohlfart T. Applying public participation GIS (PPGIS) to inform and manage visitor conflict along multi-use trails. Journal of Sustainable Tourism. 2017 doi: 10.1080/09669582.2017.1360315. [DOI] [Google Scholar]
  54. Wood SA, Guerry A, Silver J, Lacayo M. Using social media to quantify nature-based tourism and recreation. Scientific Reports. 2013;3:2976. doi: 10.1038/srep02976. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Worboys, G., and C. Pickering. 2002. Managing the Kosciuszko alpine area: Conservation milestones and future challenges. Mountain Tourism Research Report Series: No. 3. Gold Coast, Australia: CRC.
  56. Worboys G, Lockwood M, Kothari A, Feary S, Pulsford I. Protected areas governance and management. Canberra: ANU Press; 2015. [Google Scholar]
  57. Zielstra D, Hochmair HH. Positional accuracy analysis of Flickr and Panoramio images for selected world regions. Journal of Spatial Science. 2013;58:251–273. doi: 10.1080/14498596.2013.801331. [DOI] [Google Scholar]

Articles from Ambio are provided here courtesy of Springer

RESOURCES