Quantifying tourism booms and the increasing footprint in the Arctic with social media data.

Arctic tourism has rapidly increased in the past two decades. We used social media data to examine localized tourism booms and quantify the spatial expansion of the Arctic tourism footprint. We extracted geotagged locations from over 800,000 photos on Flickr and mapped these across space and time. We critically examine the use of social media as a data source in data-poor regions, and find that while social media data is not suitable as an early warning system of tourism growth in less visited parts of the world, it can be used to map changes at large spatial scales. Our results show that the footprint of summer tourism quadrupled and winter tourism increased by over 600% between 2006 and 2016, although large areas of the Arctic remain untouched by tourism. This rapid increase in the tourism footprint raises concerns about the impacts and sustainability of tourism on Arctic ecosystems and communities. This boom is set to continue, as new parts of the Arctic are being opened to tourism by melting sea ice, new airports and continued promotion of the Arctic as a 'last chance to see' destination. Arctic societies face complex decisions about whether this ongoing growth is socially and environmentally sustainable.

Introduction

The number of tourists visiting the Arctic has increased dramatically over the past two decades [1,2], reflecting a rise in tourism globally over the past 50 years [3]. While this could bring alternative livelihoods to Arctic communities, concerns over the social and environmental sustainability of the rate and scale of the tourism boom are growing across the Arctic [4-6]. Tourism can have both positive and negative impacts on the natural environment and on host communities. Direct effects of tourism (e.g. transporting, accommodating, and feeding tourists) and the indirect socioeconomic change brought about by the tourism industry (e.g. influx of seasonal workers) drive increased habitat loss [7], resource use, and carbon emissions [8] across the world, in addition to the localized consequences of nature-based tourism and recreation activities on the natural environment, such as injury, death, or disturbance of wildlife or damage to vegetation [9]. Understanding, at a local scale, how and where tourism booms are distributed across landscapes is crucial for conserving Arctic environments and for managing impacts on host communities.

A major challenge for planning and managing sustainable tourism growth in the Arctic lies in the difficulties of mapping where tourists go and how they use landscapes and ecosystems. Data on spatial visitation and trends is sparse in the Arctic. While statistics on hotel stays and transport use are now commonly collected by government and tourism management organizations, data on where tourists go during the day and what they do there is rarely collected.

Social media provides a useful source of information on tourist visitation patterns to better pinpoint the needs of tourists and target actions to manage tourism impacts [10]. Passively crowdsourced and high resolution information from social media (volunteered geographic information; VGI) can be used to rapidly generate maps of the multiple destinations visited by tourists across large areas and over time [11]. Social media data has been well demonstrated to be useful for mapping and monitoring at a range of spatial scales across the world. Such data has been used to map the distribution of tourists in protected areas [12-14], and within cities [11] and can be used to inform a variety of aspects of tourism and landscape management, including to identify peaks of visitation to attractions [15,16], map environmental impacts [17] and to estimate the landscape values, nature-based experiences and ecosystem services appreciated by tourists [18-20]. Most of these studies use the social media platform, Flickr, which has been shown to correlate well with visitor statistics at different scales [12,15,16,21].

Similar to many nature-based tourism destinations in developing countries, parts of the Arctic in Scandinavia, Iceland, Faroe Islands and Alaska have experienced unprecedented growth in the number of tourists in recent years, and Greenland, Canada and the Russian Arctic are likely to be the next tourism frontiers [22,23]. Rapid and localized booms in tourism can overwhelm local capacity (and desire) to host visitors, particularly in small, remote communities such as those found in the Arctic and many parts of the developing world [23-26]. The ability to rapidly identify sites in the early stages of a boom would support better adaptation and planning to pre-emptively address many of the sustainability challenges brought about by rapid increases in nature-based tourism. It would allow local communities and national tourism organizations to proactively direct resources to sites where special management such as provisioning of restrooms and waste disposal, better signage, safety and disaster management, parking, and trail maintenance may be required. Social media data has been demonstrated as an early-warning system for rapidly detecting booms and busts in such diverse applications as disaster management [27,28] and disease control [29]. Methods for such ‘event detection’ are rapidly evolving [30,31]. These methods rely on large amounts of high temporal resolution data (‘big data’). The suitability and limitations of social media data for detecting events in data-sparse regions has yet to be tested. We investigate whether Flickr data can be used as an early warning system to detect localized booms in tourism in the Arctic and similar data-deficient regions.

Tourism growth can influence the spatial use of landscapes in various ways. Here, we demonstrate how the tourist footprint on Arctic landscapes has changed over the past 14 years at a pan-Arctic scale, and examine the management implications of those changes. We test two hypotheses drawn from theories of tourism and economic geography [32,33] 1) that tourists visit the same sites through time (overall footprint is unchanged but magnitude of use at each site has increased) 2) that tourists have spread throughout the landscape (overall footprint increased but magnitude of use at any one site is constant). These two patterns of tourism growth have very different implications for both the social and environmental sustainability of tourism and the satisfaction of the tourist experience. For instance, if tourists avoid busy areas and self-segregate across landscapes [34], environmental and social impacts will be more widespread but lower intensity. Alternately, the negative impacts of tourism can be localized by channeling tourists into high use sites and away from sensitive communities and ecosystems. We examine seasonal differences in the spatial patterns of Arctic visitation, and explore how infrastructure such as roads, airports and ports influences these patterns, with a view to informing how upcoming infrastructure development could contribute to tourist growth.

Materials and methods

Extraction of data from Flickr

We first extracted geotagged and publicly shared photo metadata for over 2 million photos from Flickr (www.flickr.com) for the region north of latitude 60°N. Photo metadata included location and date that each photo was taken, user id (key coded by Flickr), image URL, Flickr- and user- generated image tags, and user-generated image title. Data was extracted from the Flickr API (https://www.flickr.com/services/api/) on 4 December 2017. Due to an issue with the data download we re-extracted photos for Iceland (bounded by -27° to -12° longitude and 62° to 68° latitude) on 11 January 2018. Hourly metrics of the number of photos uploaded to Flickr globally between to January 1 2004 and December 31 2017 were obtained from the Flickr API on 07 May 2018. We used the R package ‘flickRgeo’ [35] which provides an R wrapper for the Flickr API.

For the purposes of this study we define our study region, ‘Arctic’, as the region bounded by the Arctic Council AMAP boundaries [36] and confined to areas north of latitude 60°N. We constrain the study region using an environmental rather than political definition as the study is focused primarily on impacts on Arctic landscapes. We excluded photos from the extracted dataset that were taken outside this study region. We also excluded photos that were missing urls or geotag coordinates, had null coordinates (0,0), and photos taken prior to January 1 2004, or after December 31 2017. We excluded photos by users with only 1 or 2 photos within the study region as they are likely to represent people who are just trialing Flickr by uploading a random photo rather than a photo representing a genuine tourist. These ‘test users’ account for approximately 36% of users in the Flickr dataset but just 0.95% of photos. Further details on the choice of 2 photos as a threshold for exclusion are included in S1 Appendix. The final dataset contained a total of 805,684 geotagged photos with metadata from 13,596 unique users.

Mapping the seasonal distribution of Arctic tourism

To map the relative intensity of visitation across the Arctic in summer and winter, we first created square spatial grids (rasters) at 10km and 100km resolutions. We then calculated the photo-unit-days in each grid cell for summer and for winter aggregating data for each season across all years (2004–2017). We defined the months of May to October as ‘summer’ and November through April (of the following year) as ‘winter’ (e.g. “winter 2016” includes the months November and December in 2016 and January through April in 2017). Photo-unit-days (PUD) is an established metric of tourism visitation [15] that accounts for the biases in social media data introduced by differences in the number of photographs uploaded by different users. For example, three PUD can represent either a site visited on three separate days throughout the year by a single person, or by three separate people on a single day. This is the conventional approach for working with this type of social media data because it corresponds to empirical user data collected by tourism sites that are often based on daily user access fees. For example, if three users access a park with daily user access fees on the same day, or a single user accesses that park on 3 separate days, both visitation scenarios appear identical in terms of empirical visitation rates (i.e. fees collected) as well as PUD.

Pan-Arctic trends in the footprint of Arctic tourism over time

To quantify the ‘footprint’ of Arctic tourism (the percentage of the Arctic visited by tourists), we created a hexagonal spatial grid covering the entire Arctic with a 5km diameter resolution. We chose this resolution after examining the trade-off between commission errors and computing efficiency (S2 Appendix). We allocated each cell a value of 1 (visited) if any user had taken a photograph in that cell in a given year (2004–2017) and season (summer, winter), and 0 (unvisited) if not. The number of people using Flickr globally changed over time as the popularity of the platform waxed and waned. Relying on raw metrics of the number of Arctic Flickr users, photographs or photo-unit-days will thus result in biased estimates of tourism trends across time. The global and Arctic trends in Flickr use across time and the number of photos sampled in each year can be found in S3 Appendix. We calculated the footprint in three ways. Firstly, the Uncorrected Arctic footprint uses all available Flickr records (566205 summer; 228667 winter) and shows the full extent of Flickr users’ footprint, but includes the bias from the global change in Flickr usage between 2004 and 2017. The Global-bias corrected subsample removes the global pattern of Flickr usage to represent a less source-biased view of the rise in the footprint of tourism in the Arctic. This was done by selecting a random sample of Arctic photos based on the change in number of photos submitted to Flickr globally using 2004 values as a baseline. For example, the year with the lowest global usage (2004), we kept all the photo records. If the global Flickr usage doubled in a particular year relative to 2004, we sampled half of the available records for that particular year. The number of records sampled for each month can be found in Table S3 Appendix (total 36546 summer records, 12262 winter). Finally, the Equal sample size also removes the global bias and provides a measure of the change in the relative footprint per-tourist across time. This was done by randomly selecting an equal number of photographs for each year from which to calculate the footprint (total 15652 records summer, 4018 records winter). Numbers for 2017 should be treated with caution and are likely underestimates as we extracted data from the Flickr API on 4 December 2017 and the average lag time between photos being taken and uploaded is 2 weeks, but can be longer [13].

Modelling the influence of accessibility on seasonal patterns of Arctic landscape use

In order identify the effect of accessibility on the presence of tourists in different seasons, we first divided the study region into a hexagonal spatial grid with a 10km diameter resolution. We chose to use a lower resolution than that used in the footprint analysis above as the large number of cells at 5km made the models computationally intractable. We then calculated footprint in each cell for summer and for winter by allocating these cells a value of 1 if any photograph had been taken in that cell in that season (at any time from 2004 and 2017), and 0 if not, similar to the Uncorrected Arctic footprint described above. We removed cells that fell within Russia due to sparse coverage of this region by Flickr data, and from marine areas. We then modelled the footprint against variables describing the accessibility of each cell: log of distance to airports, log distance to ports, log distance to populated areas, log distance to road, the square root of the length of road within a grid cell, whether the cell overlapped any protected area, and country as a fixed effect. We chose these accessibility variables after examining a wider set of candidate variables for correlation. The models took the form of a binomial generalized additive model with logit link of footprint as a response variable. In order to account for spatial autocorrelation we included a fitted thin-plate spline on the variables latitude and longitude of each cell centroid. Smooths, intercept, slope and confidence intervals were estimated by a restricted maximum likelihood (REML) estimator and methods for large datasets (‘bam’ function in R’s mgcv package). The model formula is: where f,f are fitted thin-plate spline smooth functions on the variables latitude and longitude of each cell centroid, airport is the distance to airports, port is the distance to ports, populated is the distance to populated areas, road is the distance to road, road is the length of road within a grid cell, and PA is whether the cell overlapped any protected area (0 if false, 1 if true). Random effects are represented by β’s and the fixed effect is represented by ℽ.

The location of airports, ports and populated areas, and country boundaries were extracted from Natural Earth (www.naturalearthdata.com) using the R package ‘rnaturalearth’ [37]. Roads were extracted from Global Roads Inventory Project [38]. Protected area boundaries were drawn from CAFF [39] and supplemented with data from Protected Planet [40]. We ran separate models for summer and winter as the presence of snow and ice limits access to rural areas in winter. The summer model had 90,750 cells with no photos, and 6,554 with photos. The winter model had 93,990 cells with no photos, and 3,314 with photos.

Local trends (booms and busts) in Arctic tourism

We investigated the suitability of Flickr data to be used to identify local booms and busts in tourism in the Arctic, a data deficient region. We first divided the landscape into a square grid cells and calculated the photo-unit-days in each cell for each year. We then performed linear regression modelling to identify trends in PUD in each cell between 2012 and 2017 and test their statistical significance. We ran models for cell diameters ranging from 500m to 100km to examine the effect of data availability on trend detection. We modelled only cells that were visited in at least two of the six years between 2012 and 2017. This timeframe was chosen as global Flickr usage remained reasonably constant during this period (Fig A in S3 Appendix).

Unless otherwise stated, all analysis was conducted in R version 3.4.2 [41] using the ‘tidyverse’ [42], ‘sf’ [43], and ‘mgcv’ [44] packages. All spatial data was projected to North Pole Azimuthal Lambert equal area (EPSG:102017) for analysis. Code associated with the project is available at doi:10.18710/QEOFPY.

Results

Increase in Arctic summer and winter tourism footprint

We find that the overall footprint of tourism and area used per tourist has increased since 2004 (Fig 1). In the 10 years between 2006 and 2016, the uncorrected footprint increased from 0.066% to 0.385% of the Arctic in summer and 0.015% to 0.173% in winter. After correcting these figures to account for the increased proportion of tourists captured in the analysis over that time (i.e. accounting for the rise in Flickr use globally), the footprint increased by 374% in summer (0.029% of Arctic in 2006, 0.109% in 2016) and 634% in winter (0.006% of Arctic in 2006, 0.036% in 2016). The relative-footprint-per-tourist also increased over that period (summer: 0.028% of Arctic in 2006 to 0.043% in 2016; winter: 0.007% in 2006 to 0.012% in 2016). We caution that this metric is not the absolute per-tourist footprint, rather it should be interpreted as a relative indication of how the footprint of a fixed (and unquantified) number of tourists has changed across time. These estimates are robust to uncertainty introduced by random sampling in the Global-bias corrected and Equal sample size methods (S4 Appendix).

Fig 1

Footprint of Arctic tourism.

The total footprint of Arctic tourism measured from Flickr data increased between 2004 and 2017 (Uncorrected Arctic footprint, darker blue), even after adjusting for the global rise in Flickr use during this period (Global-bias corrected, green). The relative footprint per tourist (Equal sample size, pale blue) also increased slightly over this time. Similar trends are seen in summer and winter, though the tourism footprint in winter is approximately half the magnitude of that in summer. The footprint is defined as the percentage of 5 km hexagonal grid cells within the Arctic region visited by at least one Flickr user per year. The 2017 decline should be interpreted with caution as it may in part be an artefact of the timing of our data download and the lag between photos being taken and their being uploaded to Flickr.

Growth in Arctic visitation over time

The total number of photos on Flickr increased between 2004 and 2017, both globally and in the Arctic. Global uploads of photos to Flickr steadily increased between 2004 and 2008, slowed between 2008 and 2012 before a slight upsurge in 2012, plateaued between 2013 and 2015 and has declined slightly since then (Fig A in S3 Appendix). The Arctic represents an increasing share of Flickr’s yearly photo traffic (Table in S3 Appendix). The number of photos uploaded in the Arctic shows an exponential growth between 2004 and 2013, and has remained roughly steady since then (Fig B in S3 Appendix), with this trend overlaid on a seasonal trend in visitation (Fig C in S3 Appendix). Across the Arctic, July and August were the most popular months to photograph the Arctic (Fig D in S3 Appendix). Nonetheless, a large number of users visited during the Arctic winter (Table 1; 29.8% of all photos; 40,272 photo-unit-days, 53.3% of summer PUD). At the extreme, visitation in Greenland is concentrated in the summer months (91% of photos). In contrast to the rest of the Arctic, visitation in northern Finland peaks in winter (61.3% of photos). Winter and Christmas are a key part of the branding of tourism to northern Finland, and places such as ‘Santa Claus’s village’ in Rovaniemi, Lappland, attract the majority of the visitors to the region (Fig 2). These metrics of Arctic visitation derived from Flickr show good agreement with official metrics of tourism visitation (S1 Table), consistent with previous research on social media data [12,15,21].

Table 1

Seasonal variation in Arctic visitation, estimated from Flickr data (2004 to 2017).

	Overall number of photos (% of all photos)	Summer (% of all photos for that region)	Winter (% of all photos for that region)	Number of photo-unit-days (PUD)	Summer (% of total PUD for that region)	Winter (% of total PUD for that region)
Whole Arctic	805,684	70.2	29.8	115, 775	65.2	34.8
Iceland	377,817 (46.9)	70.0	30.0	51,500	66.4	33.6
Alaskan Arctic	150,325 (18.7)	76.0	24.0	21,903	67.0	33.0
Norwegian Arctic	122,387 (15.2)	68.2	31.8	20,779	63.7	36.3
Canadian Arctic	43,466 (5.4)	68.4	31.6	7,713	66.8	33.2
Finnish Arctic	22,164 (2.8)	38.7	61.3	3,716	43.2	56.8
Faroe Islands	20,259 (2.5)	78.1	21.9	2,226	68.0	32.0
Greenland	19,944 (2.5)	91.0	9.0	1,952	78.8	21.2
Swedish Arctic	18,303 (2.3)	53.5	46.5	2,809	56.9	43.1
Svalbard & Jan Mayen	16,282 (2.0)	71.3	28.7	1,321	72.6	27.4
Russian Arctic	13,321 (1.7)	64.2	35.8	2,228	60.7	39.3

Fig 2

Seasonal maps of Arctic tourism (2004–2017) displayed at 10km resolution.

The guide maps (right) are displayed at 100km resolution. Spatial patterns of tourism are strongly governed by air, road and sea access, with few tourists venturing far from populated areas in winter. A photo-unit-day value of 14 corresponds to one Flickr user visiting the cell per year. Country borders are modified from Natural Earth CC PD.

Flickr data not suitable as early-warning system

We investigated the potential for Flickr data to be used to identify local trends in tourism over time. The Arctic is a data-poor region and we found that only a small number of places were photographed more than once per year (Fig in S2 Appendix). Only 22 10km grid cells were visited by Flickr users more than 50 times (i.e. once a week) in 2017. Although annual tourism growth is documented as ranging from 5–20% across the Arctic, due to these data limitations regression models of trends in visitation between 2012 and 2017 were able to detect significant trends at just a handful of sites (S1 Fig). Most sites lacked sufficient data to detect trends even when aggregated to 10km grid cells, a relatively large scale compared to the scale required for management decisions.

Spatial patterns of Arctic visitation change across the seasons

The spatial pattern of tourists’ landscape use differed between summer and winter. Flickr users ventured further north and into marine areas to a greater extent during the summer months (Fig 2). Visitation was influenced by access and often, though not always, concentrated in recognized tourism hotspots (Fig 2). The main hotspots of tourism fall along coastal roads in Iceland, in the fjords and islands of northern Norway, and in protected areas and along roads in North America (Fig 2). We note that although the size of the Alaskan tourism market eclipses that of the rest of the Arctic, including Iceland, few cruises travel further north than Anchorage (61.2°N), and the majority of this tourism thus falls outside our study region which is bounded to the south at 60°N.

Accessibility drives Arctic visitation

We found that accessibility has a significant effect on the distribution of Flickr users throughout the Arctic, with the presence of tourists decreasing rapidly as distance from transport infrastructure and populated areas increases, and increasing with the length of road in a given cell (Table 2). The summer accessibility model explained 47.3% of the deviance in the presence of tourists (adjusted R2 0.448, AIC 25347). The winter accessibility model explained 51.4% of the deviance in the presence of tourists (adjusted R2 = 0.436, AIC of 14117). The variable square root of length of road in cell had the largest explanatory power of the accessibility variables (model without this variable had adjusted R2 0.409 summer, 0.393 winter, deviance explained 44.7% summer, 48.8% winter). The protected area term of the winter model was significant (p = 0.000670) in the full model, but not significant in more parsimonious models. Removal of this term only slightly decreased the deviance explained (ΔAIC = 9, Δdf = -1.0239, Δdeviance = -10.706, Pr(>Chi) = 0.001117), indicating that tourists were no more or less likely to visit protected areas than non-protected areas in winter. All other terms were significant in both models. We removed the variable that had the least explanatory power, log distance to port, from the summer model, and the winter model without the protected area term. This reduced the model fit of the summer model (ΔAIC = 145, Δdf = -1.0881, Δdeviance = -146.57, Pr(>Chi) = 2.066x10-16) but had minimal effect on the winter model (ΔAIC = 25, Δdf = -1.1028, Δdeviance = -27.406), though the large number of data points in the model meant that this variable was significant at 95% confidence (2.02x10-7). Removal of all other terms substantially decreased the deviance explained by the models. Visual examination of model residuals did not reveal any residual spatial autocorrelation in any of the models. Plots of model fit and partial plots of model variables are included in S5 Appendix.

Table 2

Model coefficients for binomial generalized additive models of the effect of accessibility on the tourist footprint in summer and winter in the Arctic.

The intercept represents Norway, unprotected. The protected area term was not included in the winter model.

	Summer			Winter
Variable	Estimate of coefficient (±standard error)	z-value	Pr(>|z|)	Estimate of coefficient (±standard error)	z-value	Pr(>|z|)
(Intercept)	8.720 ±0.619	14.094	<2.00 x 10−16	7.830 ±0.925	8.468	<2.00 x 10−16
Norway
Canada	-5.707 ±0.632	-9.035	<2.00 x 10−16	-5.690 ±0.961	-5.923	3.17x 10−9
Finland	-0.474 ±0.159	-2.988	0.0028	0.979 ±0.171	5.728	1.01 x 10−8
Faroe Islands	5.802 ±22.647	0.256	0.798	-5.557 ±0.970	-5.731	1.00 x 10−8
Greenland	-5.167 ±0.590	-8.757	<2.00 x 10−16	-4.805 ±0.893	-5.381	7.42 x 10−8
Iceland	-4.082 ±0.624	-6.540	6.15 x 10−11	-3.712 ±0.914	-4.063	4.85 x 10−5
Svalbard & Jan Mayen	-0.059 ±0.267	-0.221	0.825	-1.524 ±0.390	-3.939	8.18 x 10−5
Sweden	-0.673 ±0.134	-5.007	5.52 x 10−07	0.022 ±0.150	0.147	0.883
USA (Alaska)	-4.682 ±0.630	-7.433	1.06 x 10−13	-5.647 ±0.954	-5.917	3.28 x 10−9
Protected area TRUE	0.841 ±0.042	19.919	<2.00 x 10−16
Square root of length of road in cell	0.805 ±0.026	31.326	<2.00 x 10−16	0.574 ±0.023	24.717	<2.00 x 10−16
Log distance to road	-0.278 ±0.019	-14.912	<2.00 x 10−16	-0.278 ±0.025	-11.217	<2.00 x 10−16
Log distance to airports	-0.220 ±0.027	-8.135	4.13 x 10−16	-0.337 ±0.034	-9.855	<2.00 x 10−16
Log distance to ports	-0.381 ±0.031	-12.481	<2.00 x 10−16	-0.229 ±0.038	-5.970	2.37 x 10−9
Log distance to populated places	-0.532 ±0.022	-24.243	<2.00 x 10−16	-0.603 ±0.028	-21.244	<2.00 x 10−16

Discussion

The footprint of tourism on the Arctic environment has almost quadrupled over the past decade, from 0.03% of the Arctic in summer 2006 to 0.11% of the Arctic in summer 2016 (Fig 1), and the winter footprint has increased by over 600%. Despite this dramatic expansion, large areas of the Arctic still remain free from tourists (Fig 2). Arctic tourists tend to congregate in a handful of highly visited sites (Fig 2) with a long tail of places that are visited only occasionally, following the power law seen in other parts of the world [45]. The recent tourism boom across the Arctic has led to widespread concerns over the effects of tourism on Arctic ecosystems and communities, and the sustainability of Arctic tourism [4,5]. Arctic tourism is often marketed around ideas of pristine, untouched nature, and the overall growth in tourist numbers and the concurrent increase in infrastructure to support that growth presents challenges for maintaining environmental and social sustainability. There are indications that many popular sites are reaching capacity, and that tourists are beginning to experience disappointment and frustration around the large number of visitors present [46]. The footprint per tourist has also increased (Fig 1), indicating that tourists are now visiting a wider variety of places. This may be either due to self-segregation by tourists seeking an undisturbed experience [46], or the marketing of a wider variety of tourist attractions and nature-based activities as Arctic tourism matured over the past decade [47].

These patterns of visitation present both advantages for the management of tourists and challenges for the sharing of the economic benefits of the recent and ongoing Arctic tourism boom. Management of tourists and their impacts is often easier where they congregate in a few small areas as resources can be allocated to these high priority areas bringing economies of scale. This is particularly important in small, rural communities where both human and financial capacity to manage tourist impacts can be limited [24]. Though environmental impacts can be locally high in these well-visited areas, requiring thoughtful management, aggregation of tourists in a few small areas can reduce the impacts of tourism across the wider landscape and channel resources into efficient management at the high use sites to sustain tourist satisfaction despite crowding. Channeling visitors into such ‘sacrificial sites’ may have net benefits at the landscape scale in places like the Arctic, where wildlife and vegetation are highly sensitive to disturbance and take a long time to recover from human impacts.

The tourism footprint is strongly associated with access, and is particularly dependent on the presence of roads and airports. This is not surprising, but important to keep in mind when planning for growth in tourism. The proposed construction of three transatlantic airports in Greenland has a high potential of boosting tourism in host communities that do not have sufficient capacity to sustainably manage this growth. The sustainability of tourism in Greenland, and similar developing places around the world depends heavily on building community capacity that sustains natural resources and local culture, and that protects vulnerable sites and species from the expansion of tourists into new locations [6]. The spatial extent of the winter footprint of tourism is about 40% lower than that in summer, with tourists gathering closer to airports and towns. Managers therefore need to be especially aware of the expansion of tourists into vulnerable sites and the greater use of protected areas in the summer season. Ports have only minimal influence on the presence of tourists on land.

While social media data may be useful for rapidly detecting localized booms in tourism in highly visited regions [48], our analysis indicates that Flickr data is of limited use for identifying local tourism booms in data deficient regions such as in the Arctic. Low rates of visitation across most of the Arctic combined with the small proportion of visitors that use Flickr [15] means that just a handful of Arctic locations are visited by Flickr users more than once a month (Fig C in S1 Fig). One of the few regions where we detected statistically significant increases in visitation was the Lofoten islands of northern Norway. There, our analysis confirmed qualitative trends already noticed by tourism agencies in the region. Twitter has a higher user base and may be a better source of fine-scale temporal data in the parts of the Arctic with high-speed mobile data coverage [18,49,50]. Changes in ownership of the social media platforms and changes to data access rules means social media data from other suitable platforms such as Panoramio and Instagram were no longer freely available to academic researchers at the time of analysis. Quantitative analysis of social media data requires specialized computing and technical skills that are not normally available to local tourism management agencies. Maintaining dialogue between tourism management bodies and local communities and tour operators remains the most pragmatic way to detect and respond to fine-scale tourism trends in areas where data and technical capacity are limited.

Conclusions

The recent and rapid increase in the footprint of tourists on the Arctic that we document here is concerning. Upcoming investments in transport infrastructure in places like Greenland and the promotion of remote areas of the Arctic as tourist destinations, such as Franz Josef in Russia, will drive a further expansion of the tourist footprint in this unique part of the world. Destinations are also increasingly been marketed as ‘last chance tourism’ attracting visitors to venture into previously unexplored areas to experience Arctic ecosystems and species at risk of disappearing [4,51]. For instance, in Hudson Bay in Canada the small community of Churchill where polar bears spend increasingly more time on shore due to climate change, have experienced a rapid influx of tourists [4]. Wildlife viewing of vulnerable species, such as polar bears, narwhals and beluga whales, is putting additional pressure on species threatened by climate change [52,53]. Our high resolution and seasonal maps of Arctic tourism allow tourist management bodies and environmental organizations to pinpoint the places visited by tourists and the relative magnitude of visitation across the Arctic and to detect landscape-wide trends in visitation that need to be managed. Strategic and thoughtful assessment of whether this ongoing growth in Arctic tourism is sustainable or desirable for Arctic ecosystems and communities is urgently required.

Sensitivity analysis of photo exclusion threshold.

(PDF)

Click here for additional data file.

Sensitivity analysis of resolution.

(PDF)

Click here for additional data file.

Global and Arctic Flickr trends.

(PDF)

Click here for additional data file.

Uncertainty around footprint estimates.

(PDF)

Click here for additional data file.

GAM model performance.

(PDF)

Click here for additional data file.

Annual maps of tourism growth.

(PDF)

Click here for additional data file.

Comparison with official visitor statistics.

(PDF)

Click here for additional data file.

18 Aug 2019

PONE-D-19-16929

Quantifying tourism booms and the increasing footprint in the Arctic with social media data

PLOS ONE

Dear Dr Runge,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Both reviewers see the merits of the study reported in the manuscript. I would encourage you to address their comments in detail.

We would appreciate receiving your revised manuscript by Sep 27 2019 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Wenwu Tang

Academic Editor

PLOS ONE

Journal Requirements:

1. When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Please add a statement in the revised manuscript confirming you complied by the Terms & conditions of Flickr when gathering the data for the study.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The current study attempts to explore the spatiotemporal patterns of visitations in the Arctic with social media data. This work is an interesting topic, which is supposed to benefit the tourist management and ecosystem conservation planning.

However, due to the poorly-organized methodology and some other concerns, I recommended major revision of this manuscript. The comments are listed in the following.

(1) My major concern is about the scaling effects on the results in this study. In the context of sparse data, how does the spatial resolution affect the results, or the trend of the footprint? did the authors attempt to examine the scaling effect?

(2) Line 119-142, the sample size is the key to representative samples, so what is the specific sample size in three ways to calculate the footprint? The authors selected an random sampling method over time, what about the spatial pattern of the sampled photographs? The authors may examine the effects of the spatial pattern of the sampled records on the analysis outcomes.

(3) Line 155, how did the authors examine the correlations of candidate variables and seasonal patterns of the footprint? Does the multicollinearity exist?

(4) Line 156-160, data collection and processing should be separated from the methodology part.

(5) Line 161-163, a mathematical model is highly recommended in this part to describe the relationship between the presence of tourists and the accessibility factors, I think it will make the model description more clear. Actually, the methodology section should emphasize the model formulation, not only the analysis procedures. Besides, an analysis flowchart will make this part flow more logically.

(6) The language of this study should be further improved. Some spelling and grammar mistakes should be carefully corrected, for example, Line 144:”in order identify”, Line 270: the format of Table 2 is not consistent with Table 1, Line 328: the format of “Conclusion”.

Reviewer #2: This paper leverages social media data(geotagged flickr photos) to quantify the tourism booms and visitation patterns in the Arctic area. Their results indicate that social media cannot be used as an early warning system of tourism growth in the study area but is able to capture the growth and seasonal patterns of the visitations. Spatial statistics and modelling is used in the analysis.

I think this is a high quality research and very well-written paper The structure is clear, logic is smooth, and language is easy to understand and almost free of typos or syntax errors. Enough details are given for both data collection and processing as well as the steps of the analysis, which make it possible to reproduce the findings. I also appreciate the authors cautions in interpreting some of the results and warn readers about the potential bias etc. This paper is close to be ready for publication after addressing some minor issues as below:

1. Line 102 to 105, why only exclude those with only 1 or 2 photos? People with 3 photos are considered as tourist not test users? More explanation about the rational of setting these numbers is needed here. Could you show the distribution diagram of the number of users having number of photos? (I guess it should show a long-tail diagram?)

2. I am not convinced that PUD is a good measure here. As authors stated "three PUD can represent either a site visited on three separate days throughout the year by a single person, or by three separate people on a single day.".I don't think "a site visited on three separate days throughout the year by a single person" and "a site visited by three separate people on a single day" indicates the same visitation activity. I would like to see some justification of this in the revision.

3. For spatial aggregation, the authors used both "hexagonal spatial grid" and "square grid", why not only use one type to make it consistent?

4. Line 144: "In order" change to "In order to".

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

23 Oct 2019

Response to Review Comments to the Author

Reviewer #1: The current study attempts to explore the spatiotemporal patterns of visitations in the Arctic with social media data. This work is an interesting topic, which is supposed to benefit the tourist management and ecosystem conservation planning.

However, due to the poorly-organized methodology and some other concerns, I recommended major revision of this manuscript. The comments are listed in the following.

Response: We include detailed responses to each of the reviewer’s comments below.

(1) My major concern is about the scaling effects on the results in this study. In the context of sparse data, how does the spatial resolution affect the results, or the trend of the footprint? did the authors attempt to examine the scaling effect?

Response: Yes, we examined the scaling effect and this is detailed in S2 Appendix. Our sensitivity analysis indicated that the footprint remained reasonably constant up to approximately 25 km2 cell area. Thus for the footprint analysis, 5 km diameter hexagons (21.6 km2 cell area) were an appropriate compromise between overestimating the footprint (commission errors) while keeping the number of cells low enough that computations could be undertaken with the memory available to us. For temporal trend estimates, sensitivity analysis indicated that 10 km diameter hexagonal cells (86.6 km2 cell area) seems a good compromise between providing spatial resolution at a scale relevant for management and having sufficient data in each cell to analyse trends.

(2) Line 119-142, the sample size is the key to representative samples, so what is the specific sample size in three ways to calculate the footprint? The authors selected an random sampling method over time, what about the spatial pattern of the sampled photographs? The authors may examine the effects of the spatial pattern of the sampled records on the analysis outcomes.

Response: We now include the total sample size for each method in the manuscript which we previously included by month and season in Table S2 (now Table S3). The reviewer is correct, random subsamples used in two of the three methods to calculate the footprint will generate differences in the spatial patterning between random draws. We performed a sensitivity analysis of the uncertainty introduced by random sampling which we now include as Appendix S4. The results indicate that our findings are robust to the uncertainty introduced by sampling.

(3) Line 155, how did the authors examine the correlations of candidate variables and seasonal patterns of the footprint? Does the multicollinearity exist?

Response: As we state on L162 we retained the specified set of variables which showed low multicollinearity, after examining a wider set of candidate variables for collinearity.

(4) Line 156-160, data collection and processing should be separated from the methodology part.

Response: We respectfully disagree, and believe that summary of the methods used in data collection and processing is best described within the methodology section.

(5) Line 161-163, a mathematical model is highly recommended in this part to describe the relationship between the presence of tourists and the accessibility factors, I think it will make the model description more clear. Actually, the methodology section should emphasize the model formulation, not only the analysis procedures. Besides, an analysis flowchart will make this part flow more logically.

Response: The model follows the general formulation below, which we now include in the manuscript on L169:

g(footprint) =β_0+ f_1 (long) + f_1 (lat) +β_1 〖log〗_10 (airport_dist) + β_2 〖log〗_10 (port_dist) +β_2 〖log〗_10 (popula〖ted〗_dist) + β_3 〖log〗_10 (〖road〗_dist)

+ β_4 √(〖road〗_length ) +β_5 PA +γ_6 Country +ε

where g is the logit link function, f1,f2 are smooth functions estimated by the model by REML, airportdist is the distance to airports, portdist is the distance to ports, populateddist is the distance to populated areas, roaddist is the distance to road, roadlength is the length of road within a grid cell, and PA is whether the cell overlapped any protected area (0 if false, 1 if true). Random effects are represented by β’s and the fixed effect is represented by ℽ.

(6) The language of this study should be further improved. Some spelling and grammar mistakes should be carefully corrected, for example, Line 144:”in order identify”, Line 270: the format of Table 2 is not consistent with Table 1, Line 328: the format of “Conclusion”.

Response: Done. We note that the journal has a typesetting process for tables post review. Conclusions is a subheading within Discussion as suggested in the journal guidelines, hence the different formatting.

Reviewer #2: This paper leverages social media data(geotagged flickr photos) to quantify the tourism booms and visitation patterns in the Arctic area. Their results indicate that social media cannot be used as an early warning system of tourism growth in the study area but is able to capture the growth and seasonal patterns of the visitations. Spatial statistics and modelling is used in the analysis.

I think this is a high quality research and very well-written paper The structure is clear, logic is smooth, and language is easy to understand and almost free of typos or syntax errors. Enough details are given for both data collection and processing as well as the steps of the analysis, which make it possible to reproduce the findings. I also appreciate the authors cautions in interpreting some of the results and warn readers about the potential bias etc. This paper is close to be ready for publication after addressing some minor issues as below:

Response: We thank the reviewer for their positive reception of our manuscript.

1. Line 102 to 105, why only exclude those with only 1 or 2 photos? People with 3 photos are considered as tourist not test users? More explanation about the rational of setting these numbers is needed here. Could you show the distribution diagram of the number of users having number of photos? (I guess it should show a long-tail diagram?)

Response: We examined the characteristics of different types of users during the data cleaning process we undertook before analysing this dataset. When we examined the photographs of a sample of ‘test users’ (those with 1 or 2 photos) these appeared to be ‘random snaps’ i.e. out of focus or with no discernable subject. Thus we felt justified in excluding these users. The reviewer makes a good point about why choose 2 photos as a cutoff point, and we acknowledge this threshold is somewhat arbitrary one that attempts to balance retaining quality photos without excluding too many users. People contributing 10 or fewer photos account for less than 1% of the number of photographs, but over two thirds of the users. This dropped to one third of users with 2 or fewer photos. We now include the histogram and discussion of this issue in Appendix S1. The median number of photos per user ranges across countries from 25-33 photos.

2. I am not convinced that PUD is a good measure here. As authors stated "three PUD can represent either a site visited on three separate days throughout the year by a single person, or by three separate people on a single day.".I don't think "a site visited on three separate days throughout the year by a single person" and "a site visited by three separate people on a single day" indicates the same visitation activity. I would like to see some justification of this in the revision.

Response: This is the conventional approach for working with this type of social media data because it corresponds to empirical user data collected by tourism sites that are often based on daily user access fees. For example, if three users access a park with daily user access fees on the same day, or a single user accesses that park on 3 separate days, in terms of empirical visitation rates (i.e. fees collected) as well as PUD both visitation scenarios appear identical. We now clarify this in the text on L119.

3. For spatial aggregation, the authors used both "hexagonal spatial grid" and "square grid", why not only use one type to make it consistent?

Response: The two grid types each have benefits for the two different analyses. Hexagonal grid cells minimise spatial biases introduced by gridding data, and we regarded as a better choice for the footprint analyses than square grid cells. Ideally we would have used hexagonal grid cells for both analyses, however the large number of cells meant that for the model analysis we needed to take advantage of the computational efficiencies available from raster processing tools.

4. Line 144: "In order" change to "In order to".

Response: Done

16 Dec 2019

Quantifying tourism booms and the increasing footprint in the Arctic with social media data

PONE-D-19-16929R1

Dear Dr. Runge,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Wenwu Tang

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: (No Response)

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

20 Dec 2019

PONE-D-19-16929R1

Quantifying tourism booms and the increasing footprint in the Arctic with social media data

Dear Dr. Runge:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Wenwu Tang

Academic Editor

PLOS ONE