Global coral reef related tourism is one of the most significant examples of nature-based tourism from a single ecosystem. Coral reefs attract foreign and domestic visitors and generate revenues, including foreign exchange earnings, in over 100 countries and territories. Understanding the full value of coral reefs to tourism, and the spatial distribution of these values, provides an important incentive for sustainable reef management. In the current work, global data from multiple sources, including social media and crowd-sourced datasets, were used to estimate and map two distinct components of reef value. The first component is local “reef-adjacent” value, an overarching term used to capture a range of indirect benefits from coral reefs, including provision of sandy beaches, sheltered water, food, and attractive views. The second component is “on-reef” value, directly associated with in-water activities such diving and snorkelling. Tourism values were estimated as a proportion of the total visits and spending by coastal tourists within 30 km of reefs (excluding urban areas). Reef-adjacent values were set as a fixed proportion of 10% of this expenditure. On-reef values were based on the relative abundance of dive-shops and underwater photos in different countries and territories. Maps of value assigned to specific coral reef locations show considerable spatial variability across distances of just a few kilometres. Some 30% of the world's reefs are of value in the tourism sector, with a total value estimated at nearly US$36 billion, or over 9% of all coastal tourism value in the world's coral reef countries.
Mapping the global value and distribution of coral reef tourism
MarkSpaldinga. LaurettaBurkeb.Spencer A.Woodc. JoscelyneAshpolee. JamesHutchisone. Philinezu Ermgassene. http://www.sciencedirect.com/science/article/pii/S0308597X17300635#f0015
Coral reef related tourism is an important and still fast-growing industry, providing employment and income to over 100 jurisdictions world-wide, and often generating much-needed foreign earnings [1–6]. Coastal tourism in the vicinity of coral reefs is not always benign: negative impacts can include degradation and loss of marine life through activities such as diving and snorkelling [7–9], as well as indirect impacts arising from poorly planned coastal development, including dredging, building on intertidal spaces, and increases in pollution and solid waste [10–12]. Despite these risks, tourism may be a less significant threat than fishing, land-based run-off or coral bleaching , and may even help to reduce some threats, notably over-fishing, by offering financial or social incentives for sustainable management [14–17]. Many visitors to coral reefs already have heightened environmental awareness  and reef visitation can both help to fund  and to encourage  coral reef conservation.
Much of the focus on the value and impacts of coral reef tourism has focused on the direct use of coral reefs for in-water activities. The indirect value of coral reefs in driving coastal tourism remains less well quantified, but is also important. Studies have shown the considerable importance of clear water and beach characteristics such as fine sand in influencing tourist preferences [21–23]. There is also a sizeable literature on the multiple ecosystem functions provided by coral reefs which may support tourism benefits, including the generation of fine sand beaches , the maintenance and building of islands [25,26], protection from wave erosion and storm damage , and the production of seafood . Coral reef imagery also plays an important role in marketing , while the perception of reef proximity, even for non-reef users, may be an important draw [23,30]. Finally, the health benefits associated with proximity to nature and to marine environments more generally are now increasingly realised [31,32], and are also likely to be played out in coral reef settings.
Given the broad array of economic and social benefits that coral reefs provide, there are growing attempts to build more sustainable approaches to reef-related tourism [33–37]. In large part, however, such efforts remain the target of individual operators, a few small island nations or some operators in the diving sector. The wider call for corporate social responsibility (CSR) has been slow to influence the tourism sector in general, with biodiversity conservation remaining low on the agenda even among those corporations who are engaging in CSR more broadly . Against this background, a clear, quantified and reliable understanding of the value of coral reefs for tourism in specific settings could help, by raising awareness and highlighting opportunities for strengthening coral reef conservation, both in the public sphere and in the tourism sector.
Previous efforts to quantify the value and to describe the spatial distribution of reef-related tourism at large scales have been limited. Without mapping, de Groot et al.  generated a mean value for coral reef recreation of US$96,302 ha−1 yr−1. This figure was derived from 29 studies, with a median value of US$1562 ha−1 yr−1, but ranging from zero to almost US$1.5 million ha−1 yr−1. The large mean value from this study was used in preference to the median value in a direct benefit transfer approach to all coral reefs, generating a global estimate of value of US$2.7 trillion per year, or 2.2% of all global ecosystem service values [derived from supplementary materials in 40], a figure that seems impossibly high given the spatially restricted nature of coral reef tourism. Elsewhere, Brander et al. had already drawn attention to the challenges of such extrapolation: with data from 100 separate reef recreation studies, they conducted out-of-sample value transfer tests and estimated average transfer errors of 186%, a figure they deemed “unlikely to be acceptable in most policy-making scenarios” (pg 215).
Given the challenges of developing value transfer approaches, alongside the acknowledged benefits of developing an understanding not only of global values but of the spatial distribution of such value, this work presents a novel approach to accurately quantify global reef values and to distribute these values to specific reefs at local scales. The work draws on a variety of data-sources including national-level visitor and expenditure statistics, together with locally accurate data from industry, social media and crowd-sourced datasets to support the spatial modelling of value distribution.
2. Framework and methods
In this work, coral reef related tourism is defined as the combined tourism and recreation activities that can be attributed to the presence of coral reefs. The value of such tourism is explored here using metrics of monetary value and tourist trip equivalents (a term used to account for the fact that many of the statistics are built up from fractions of total arrivals, acknowledging that reefs are only part of most destination choices). Two components of coral reef tourism are distinguished: on-reef or in situ values are derived from direct non-extractive reef uses including diving, snorkelling and glass-bottom boat tours ; reef-adjacent or ex situ values are not derived from in-water activities, but are indirectly linked to the presence of nearby reefs. The latter may include the role of reefs in generating clear calm waters and beach sand, outstanding views, fresh seafood and even their widespread use in advertising, all of which help to draw people to coral reef regions. Recreational fishing may represent an important additional value which we were unable to incorporate into our model, in large part due to the highly variable nature of reef dependency: much recreational fishing in coral reef countries targets non-reef fish in offshore waters.
Tourism was examined in the over 100 jurisdictions (countries and territories) around the world which have coral reefs. Statistics on travel and tourism to each jurisdiction were used as initial inputs. “Big data” from commercial, crowd-sourced, and social-media platforms were then used to make predictions of tourism expenditure and visitation to non-urban areas that could reasonably be linked to coral reefs. Paired, independent datasets were used to strengthen the robustness of the model, and the outputs were cross-referenced to existing studies. Finally, these tourism values were linked back and assigned to the reefs that were considered to be generating these values. Fig. 1 gives a schematic of this work, while details of the datasets and the methodological process are described below.
Jurisdictional-level tourism. Expenditure statistics were largely taken from the World Travel and Tourism Council  and represent spending for tourist and business trips, including travel and accommodation. Jurisdictions were largely countries and territories, although, due to the strong geographical differences, the US states of Hawaii and Florida were considered as distinct jurisdictions. Arrivals, largely derived from the UN World Travel Organisation (UNTWO) , include recorded overnight stays by international, cruise ship, and domestic visitors; they do not incorporate length of stay. The decision to include all travel in this initial step was driven by a desire to ensure a more complete and consistent dataset. Non-relevant travel was filtered out as outlined in the subsequent steps (below). Gaps in these data were filled from other international or national, sources (see Appendix A). Where possible, data were gathered for all years from 2008 to 2012. Local currency data were converted to historic US dollar values for 30 June of the relevant year. These values were then converted to 2013 values using the Consumer Price Index (CPI) price deflator (data.worldbank.org).
Distribution of national tourism. In order to develop an estimate of the geographic distribution of national tourism and expenditures, two independent data sources were used: the distribution of hotel rooms, and the distribution and frequency of geo-located photographs from the popular image-sharing website Flickr. Hotel rooms give an approximate measure of overnight visitor intensity. The Global Accommodation Reference Database (GARD), a commercial database, was generously made available (http://www.delta-check.de) for the coral reef nations of the world. This provided location and size (number of rooms) data for 125,498 hotels in coral reef jurisdictions. Publicly sourced photographs from Flickr include large quantities of geo-located photographs world-wide – an estimated 40–50 million have been uploaded to www.flickr.com annually since 2010. The geographic distribution of photos gives an indication of human activity and particularly leisure-based activities, and the spread and density of such photos have already been developed as a metric to quantify tourist activity . Following the same approach as that work, a consistent spatial metric of the intensity of photography was utilised. Known as photo-user-days (PUD), these represent the total number of days, across all users, that at least one photograph was taken in a given area . Average annual PUD per ~1 km gridded cell was computed for all coral reef nations and territories, including offshore waters, for the years 2005–2012.
Wood et al.  and subsequent authors [45,46] have recognised possibilities of biases in the spatial distribution of Flickr images, based on who is uploading photographs and the numbers of photographs being taken of different recreational activities. With the hotels layer the present authors were not aware of any biases, however reporting may be better for some locations than others. Both layers were independently derived, but on comparison they tracked one another relatively well (N=103 coastal states with hotels and PUDs, Pearson's r =0.899, p<0.001). Using each layer, a weighted distribution map was generated, showing the national tourism values assigned to locations based on the numbers of PUDs and, independently at this step, hotel rooms for each country or territory.
Coastal non-urban tourism near reefs. It was then necessary to filter out the tourism values that were unlikely to be reef-related from the national totals. Using weighted distribution maps described above, values were removed from three broad geographic extents: a) Non-coastal areas, defined as any areas beyond 2 km landwards and seawards of the coastline, b) Urban areas, defined as towns and cities of greater than 10,000 people , and c) Non-reef areas, defined as areas over 30 km from reefs that are unlikely to be benefitting in any substantial way from the presence of reefs . The influence of business travel, deliberately included in the initial input statistics (1, above), is likely to be considerably reduced in this stage, as a large proportion will be urban and non-coastal. For the remainder, it was considered likely that business travel to coastal resorts near coral reefs will in part be driven by the attraction that reefs provide, and therefore that it was important to include such travel.
The resulting two maps of coastal non-urban PUD and hotel-rooms were then aggregated into a single statistic for coastal non-urban tourism within 30 km of a coral reef (hereafter termed reef-coast tourism) for each country. While the correspondence between PUDs and hotel rooms was reassuring, it was felt that both layers, in different places, showed varying levels of completeness, and further that by combining both sources of information any unseen biases, or more local effects, could be reduced. To achieve this, the two layers were combined with a weighting of 2:1 towards the use of PUDs, which gave a more detailed spatial portrayal of visitor use than hotel rooms. This layer showed that 44% of coastal tourism in coral reef jurisdictions occurs within 30 km of reefs. If the distance was reduced to 10 km, that figure only dropped to 36%, suggesting that most tourism is in fact much closer to coral reefs than 30 km. Clearly not all tourism in this zone is attributable to coral reefs (see following steps), but this initial step provided a clear starting point for understanding the proportion of coastal tourism potentially attributable to coral reefs.
Reef-adjacent values. Many tourists do not take part in on-reef activities such as diving, snorkelling and boat-trips, but coral reefs may still play a critical role in attracting them to particular locations. While such reef-adjacent benefits are widely agreed to be key drivers of tourism in many locations, the authors were unaware of any existing models which would enable the prediction of such values across multiple jurisdictions. Many existing reef valuation studies look only at on-reef values (reef visitation, diver numbers, etc.), while others provide total values for “reef regions”, with the implicit assumption that all tourism can somehow be related to the presence of the reef. Given that the on-reef values were disaggregated separately (below), the approach taken here for these reef-adjacent values was to assign a simple estimate of 10% of the value of all reef-coast tourism to coral reefs. This is likely to be conservative: some other studies have assigned higher values, but these could also be more easily disputed. For example, Sarkis et al.  suggest that visiting the beach was a prime motivation for some 16% of visitors in Bermuda, while arguing that the beaches themselves are formed by the reefs and thus entirely coral reef dependent.
On-reef tourism values. Some of the highest tourism values from individual reefs are linked to direct, in-water uses, notably scuba diving, snorkelling and boat tours. The magnitude of on-reef tourism is influenced by a suite of natural, social and economic factors including biodiversity and reef health, ease of access to reefs, available infrastructure, history, and culture. Existing efforts to quantify on-reef tourism cover few jurisdictions and there is little consistency in methods. A novel approach was therefore devised to quantify the relative importance of on-reef tourism as a proportion of total reef-coast tourism, with help from existing studies to help calibrate these efforts (see Appendix A). This approach used two proxy measures: abundance of dive shops relative to hotel rooms and abundance of underwater photographs relative to all photographs shared on Flickr. For the first metric, data on the locations of more than 4000 dive shops world-wide were obtained from a crowd-sourced database, generously provided by www.diveboard.com. These were used to generate statistics on the number of dive shops per 1000 hotel rooms within the reef-coast region of each country. For the second metric, the collection of geo-located photographs shared on Flickr (see Method 2 above) were used to obtain a subset of images that were tagged with a keyword related to underwater recreation. Multiple search terms in eight languages were used to maximise the reach of this subset (see Appendix A). In total, the search identified over 14,500 PUDs according to the underwater images within 30 km of a coral reef. The ratio of underwater PUDs to total PUDs were then generated in the reef-coast regions of each country.
These two ratios gave independent estimates of national on-reef use intensity. Ignoring values where either dataset held zero data there was a significant positive correlation between the rank-orders of jurisdictions developed from each dataset (n=89 jurisdictions with both datasets, Pearson's r =0.632, p<0.001).
While the authors felt the overall approach was strong, they were also aware of gaps or weaknesses in some of these datasets. For example, there were four jurisdictions where the (limited) dive tourism was not represented by any dive shops. Given these concerns it was felt appropriate to use both datasets, alongside some further expert intervention. Both datasets were scaled between 1 and 100 to produce relative measures. These numbers were then compared against each other and against several external sources (see Appendix A). Overall, underwater PUDS were the more spatially detailed and sensitive of the two datasets, and were also likely to represent the full range of in-water activities, including snorkelling and glass-bottom boat tourism alongside diving. Thus, a single on-reef use intensity statistic was generated, weighted 2:1 in favour of underwater PUDs over the dive shop metric. Where the two on-reef tourism metrics were well correlated, and where there was no evidence for errors (86 of the 102 jurisdictions with tourism) this single weighted number was used, capped at 70%, as the preferred metric for the proportion of remaining reef-coast tourism which could be assigned to on-reef value.
For the remaining 16 territories known to have reef tourism, but with limited data or showing disagreement among the underwater PUD and dive-shop metrics, an equivalent number was generated through the following process (see Appendix A for further details). Where data were missing or very limited from either one of the two metrics (i.e. there were few or no underwater PUDs or dive shops), a new score was developed that was weighted towards the other metric, or used that metric alone (three territories). Where data were considered poor and were clearly divergent from a recent and comparable literature source, the numbers were replaced with that alternative (four territories). Where there was evidence, based on expert-knowledge and/or available references, that both metrics were under- or over-reporting, scores were altered to the best approximation of the authors (nine territories).
Finally, the remaining 90% of reef-coast tourism (excluding the 10% assigned to reef-adjacent values) for each jurisdiction was multiplied by the estimate for on-reef tourism estimates described above (ranging from 0% to 70%) to give an estimate of on-reef value for each jurisdiction.
Value attribution to reefs. The final stage of this work was to separately assign the reef-adjacent and the on-reef values to the reefs that are likely to be generating those values. To ascribe reef-adjacent tourism expenditures, the national dollar and tourist trip equivalent values were distributed separately following PUD densities and hotel rooms within 30 km of a coral reef. The two layers were then combined with a 2:1 weighting as above, and the values from this combined layer were then attributed to nearby reefs (up to 30 km distant), using a series of steps (see Appendix A) with a weighting to assign the greatest proportion of these values to the nearest reefs (within 5 km).