For each case study area, a search query was executed (Table 1). Query terms were based on the hashtags of the geographical name of the study areas; therefore, the post download was related to the name of the study area (e.g., Galapagos), with all downloaded posts including this name as query. Query search was limited to English, the most common language amongst tourists. This might have overlooked posts where the name of the place was in a different language. For most marine areas, this was considered irrelevant as the name of the place is not translated to other languages (e.g., Tayrona, Vamizi, Skomer). In some of the cases, the name of the place could appear in a variety of languages (e.g., Great Barrier Reef), however, the use of non-English place hashtags as queries generally retrieved a significantly lower number of posts (e.g., Gran Barrera de Coral in Spanish with 1900 posts, or Grand Barrière de Corail in French with 14 posts, while Great Barrier Reef had over 10,000 posts). In the specific case of Easter Island, we observed that the use of three particular queries was linked to a high number of posts: Easter Island and the local name Rapanui had over 10,000 posts each, and Isla de Pascua in Spanish had 8700 posts. In this case, three separate posts’ downloads were performed, and data were merged for subsequent analysis. The above, rather than a limitation of the methodological approach, demonstrates its flexibility to adapt to different data acquisition requirements.
To illustrate the most relevant information contained as part of the posts downloaded for each of the 14 areas, we selected the 150 most frequent hashtags from each dataset in order to create the network graph and represent the dominant discourse in relation to the area in question. Network graphs were delineated using eigenvector, betweenness and edge betweenness as centrality measures. Eigenvector centrality measure (hereafter Eigenvector) allows identifying those hashtags that are frequently posted with other hashtags also frequently posted, and it can be interpreted as the pairs or groups of features more frequently related to the case study by the users. Betweenness centrality (hereafter betweenness) and edge betweenness centrality (hereafter edge betweenness) provide information about clusters of hashtags that describe users’ experiences or perceptions and that connect (by means of a hashtag) to other clusters representing other types of experiences or perceptions. These high betweenness hashtags structure the general discourse about an area and their removal would fragment the network and disconnect distant concepts. Therefore, hashtags and links with high betweenness can show the discourse parallel or additional to the main discourse and their relations, allowing to identify less frequent activities or perceptions but that are equally important to understand the network as a whole.
Network centrality measures
Results indicated that network graphs captured information on distinct types of ecosystem services, for example, those based on wildlife and nature, heritage, or beach tourism. In areas such as Galapagos, central hashtags were nature, wildlife, photography, travel and adventure, evidencing a preference for wildlife and nature-based tourism. In this area, betweenness evidenced the connections between the most frequent hashtags group with other peripheric hashtags and provided a complete picture on the discourse of Galapagos’ visitors (Fig. 2). As such, nature and wildlife-based travel and photography is related with natural science concepts like evolution and endemism, and specific biotic and abiotic components like crabs and waves, altogether related with positive feelings (i.e., happy). Other areas emerging for their wildlife and nature were Skomer nature reserve, characterised by the hashtags birds (including the species Puffin), nature and wildlife photography; and Península Valdés, characterized by many locality names and by fauna, with the frequently posted hashtags’ wildlife, whales and nature funnelling most connections to other less frequent hashtags (e.g., wind, hiking, relax) and providing a full picture of the social perception on nature recreation activities, iconic fauna and positive feelings. Three networks, Sandwich Harbour, Glacier Bay and Macquarie Island also included popular hashtags related with nature, wildlife and photography; however, most hashtags had low betweenness and edge betweenness limiting the diversity of the posts (all network graphs are available at the Figshare repository, https://doi.org/10.6084/m9.figshare.13325627.v2).
Example of network graphs in Galapagos case study. In plot (A) node size represents the Eigenvector centrality and edges represent normalized strength (weighted degree). In plot (B) node size represents normalized Betweenness centrality and edges represent normalized Edge betweenness.
Regarding cultural heritage, Easter Island was characterised by popular hashtags related with Easter Island stone statues (moais) and with travel; and edge betweenness evidenced a diversity of peripherical nodes that describe other cultural elements, like design, music and food, and evidence social preferences for different cultural elements of the island, beyond the moais. Other areas reflected cultural identity by the frequent post of local names (e.g., Ytrehvaler), words related with the country’s identity (e.g., Isole Egadi) and positive feelings about this identity (e.g., Tawharanui). In Tayrona National Park network, the full discourse identified cultural identity like Kogui (indigenous culture) linked with the popular posts related with nature and summer holidays. Similarly, in Tawharanui and Isole Egadi, beach, nature and summer where the most frequent posts that, in some cases, where connected with places and activities. In these cases, and particularly in Isole Egadi and Ytrehvaler, edge betweenness allows to identify connections between places and activities, wildlife or natural structures, providing relevant information for area management and conservation.
A group of areas were appreciated by their underwater ecosystems. For Great Barrier Reef, popular hashtags were related with the coral reef: ocean, diving, underwater photography, travel, nature, coral and reef; whereas betweenness highlighted a set of hashtags related with conservation: science, sustainability, save the reef, 4 ocean (Fig. 3) and evidenced the presence of a conservationist discourse in the social media. In Toguean Island network, the frequent hashtags beach, wonderful and charming are connected to peripherical hashtags related with the sea (e.g., sea life, diving), while in Vamizi, popular hashtags were related with high-income tourism, private island, travel, luxury travel, and were connected to less frequent hashtags linked to the sea, including recreational fisheries. These last two examples illustrate differences in the benefits, and beneficiaries, provided by two popular touristic destinations.
Example of network graphs in Great Barrier Reef case study. In plot (A) node size represents the Eigenvector centrality and edges represent normalized strength (weighted degree). In plot (B) node size represents normalized Betweenness centrality and edges represent normalized Edge betweenness.
Network communities
The division of hashtags in communities allows for a more detailed exploration of the words included in the 150 most frequent hashtags selection, independently of their centrality measures, and allowed a categorisation of hashtags within cultural ecosystem services classes in each area (Table 2). Hashtags were grouped in 3 to 5 communities, with some communities relatively constant across case studies, e.g., aesthetics, wildlife and nature appreciation (Fig. 4) (all other network graphs are available at the Figshare repository, https://doi.org/10.6084/m9.figshare.13325627.v2).
Communities assessed through Fast-Greedy algorithm for the case studies Glacier Bay (A) and Tayrona (C). The node size represents the normalized Eigenvector and the colour represents the community. The colour and width of the edges represents the normalized edge strength (weighted degree).
In some of the areas, the communities were diverse in hashtag composition, for example, in Galapagos, wildlife (and related words) was distinctive of several communities, but other communities were characterised by different concepts: beach, holidays, happiness, snorkelling and diving. In Easter Island, the hashtags related with the stone statues and cultural heritage characterise one community, while the other communities include a diversity of hashtags classified under adventure, nature, underwater recreational activities; therefore, it widens the information provided by the centrality metrics. Tayrona (Fig. 4) is also a diverse network with one community characterised by hashtags like beach, summer, happiness (wellbeing), but other communities contain a diversity of hashtags like forest, hiking, indigenous and wildlife (classified in recreational, cultural heritage, nature and aesthetics; Table 2).
In some areas, the communities were not so diverse, but provided additional information on the posts. For example, in MacQuarie Island the communities highlighted iconic fauna, including several penguin species, and biodiversity conservation. In several areas, network communities informed of the iconic fauna and specific places: puffins and other bird species in Skomer; southern right whale, sealions and penguins in Península Valdés; glaciers and mountains in Glacier bay (Fig. 4); desert and dunes in Sandwich harbour. Finally, Ytrehvaler is a network characterised by many local names (in Norwegian), evidencing a national tourism, and hashtags related with scenery.
Merged network of the 14 case studies
The merged network highlighted several hashtags that act as bridges between communities of hashtags (Fig. 5). Nature, travel, photo and travel photography are key to structure the global network. However, several low eigenvector hashtags connect smaller groups: sunset and island connect the subgroups from Easter Island, Isole Egadi and Vamizi.
Global network graph including the fourteen case studies where the node size represents the Eigenvector centrality. The coloured clusters arrange the case studies to facilitate the visual identification of areas connected in the network.
From the hashtag travel photography diverges a branch that connects 7 areas through adventure; a small group of hashtags deriving from this node represent Sandwich harbour and Vamizi, connected through Africa. The hashtag ocean, connected to adventure, relates Great Barrier Reef with Tawharanui, and to wanderlust (a German expression for the desire to explore the world) that connects Península Valdés, Skomer and Macquairie Island. These three areas and Tayrona are also connected through the central hashtag travel photography, and Skomer and Macquairie Island through wildlife photography. The hashtag adventure is also connected to a group of hashtags from Galapagos that also derive to the high eigenvector hashtag nature.
The hashtag nature is key to include the fragile sub-network Ytrehvaler, and also derives to other high eigenvector hashtag, travel, that in turn, connects to the small sub-network from Glacier bay. Photo, a central hashtag related with travel, connects to paradise, that is key to integrate Toguean Island, a few hashtags from Tayrona related with the Caribbean and beach, and a group of hashtags from Peninsula Valdez related with whale watching. Some other small hashtags, that are connected to high eigenvector hashtags but are not included in any particular area are shared by many of the areas, e.g., sun, relax, landscape photography, nature lovers, sunset, sky.
Source: Ecology - nature.com
