What on-site landscape experiences attract potential visitors to a site? A Japan–Korea cross-cultural comparison

ABSTRACT In the age of big data, social media’s influence on on-site landscape experiences is growing, and the relationship between on- and off-site landscape planning and management is becoming more important. The study’s objectives were: 1) to identify scenic hotspots based on on-site visitor experiences, 2) clarify landscape preferences and attractiveness based on photographs taken at scenic hotspots, and 3) examine similarities and differences between Japanese and Korean visitors in terms of landscape preferences and attractiveness. The data were collected using geotagged visitor employed photography from 153 visitors to the Takao Quasi-National Park in Japan were used to understand off-site appreciation, and compare Japanese and Korean visitors’ experiences. Photo-based questionnaires were completed by 42 Japanese and 40 Korean respondents. We collected 1,645 geotagged photographs, and performed a GIS analysis with network-based kernel density estimation to identify scenic hotspots. The data extracted from the photo-based questionnaire were used to analyze the relationship between on-site and off-site experiences, as well as similarities and differences between the Japanese and Korean participants. The data were analyzed using means, Welch’s t-test, Wilcoxon sum-rank test, Pearson’s correlation analysis, and Kendall’s correlation analysis. The results identified nine scenic hotspots, and revealed that the photographs that the Japanese visitors preferred to take would not necessarily attract tourists, nor were they the same as the scenic spots and objects the Korean visitors found attractive. As international tourism expands, cross-cultural research on on- and off-site experiences and preferences has become increasingly important for forest landscape management and sustainable tourism.


Introduction
Interest in nature-based tourism (NbT) is growing.Forests play an essential role in providing recreation and tourism opportunities that can have a significant impact on economic, social, and environmental sustainability (Ghorbanzadeh et al. 2019;Fossgard and Fredman 2019;Haukeland et al. 2023).To manage forests and to make them more attractive to visitors, which can also contribute to NbT, it is important to understand what tourists want to visit and why (Colson et al. 2010;Bell 2010).As attractive landscapes are an important factor in NbT (Mäntymaa et al. 2021), there is a growing interest in forest landscape planning and management, as well as its scientific production of them (de Jesus França et al. 2022).
Landscape perception theory (Zube et al. 1982;Gobster et al. 2022) states that landscape perception is affected by humans, forests (environments), and their interactions.Many empirical studies have focused on forest conditions and visitor backgrounds (Daniel and Daniel TC 2001;Ueda et al. 2012;Hansen 2020), in other words, similarities and differences in the landscape perceptions based on nationality (Yang and Brown 1992;Purcell et al. 1994;Sommer 1997;Aoki 1999;Herzog et al. 2000;Fairweather and Swaffield 2002).In light of the growing popularity of international tourism, understanding the similarities and differences in landscape perceptions by nationality has become increasingly important to provide better visitor experiences (Terkenli et al. 2021).
In an area of the sociology of tourism, there is an interesting theory named "tourist gaze", which refers to the way in which tourists see and experience the places they visit are constructed socially (Urry 1990).Urry also developed the concept of the "circles of representation".It refers to a feedback loop about a destination's image projected by the tourism industry and what tourists want to see.Based on this destination's image, tourists visit the destination, take photographs, and show photographs to others, thereby reinforcing the destination's image.Some empirical studies have attempted to confirm this theoretical concept (Jenkins 2003;Garrod 2008Garrod , 2009)).Moreover, the widespread use of smartphones and the popularity of social networking services (SNS) have strengthened the influence of individual tourists on a destination's image formation process in the "circles of representation" (Stepchenkova and Zhan 2013;Konijn et al. 2016;Balomenou and Garrod 2019).Photographs taken onsite by individual tourists are viewed and evaluated via SNS, thereby serving as a source of tourism information, and impacting the landscape perceptions.A growing body of studies is using geotagged photographs uploaded to SNS to understand the places and landscapes that tourists prefer to visit (Heikinheimo et al. 2017;Bubalo et al. 2019;Barros et al. 2019;Jianrong and Zhenbin 2022).These places are analysed using GIS and are called key viewpoints (Palmer 2019) or scenic hotspots (Brown and Murtha 2019).
Hence, it is crucial for those involved in the planning and management of forest landscape to comprehend the sequence of relationships between individual tourists' site preferences, how photographs are taken, and how they are appreciated off-site.The objectives of this study are: 1) to identify scenic hotspots based on visitor's on-site experiences, 2) clarify their landscape preferences and the attractiveness of the photographs taken at scenic hotspots, and 3) clarify the similarities and differences between Japanese and Korean potential visitors to the Takao Quasi-National Park (QNP) in Japan in terms of landscape preferences and attractiveness.

Site description
The study was carried out in the Mt.Takao designated as a Takao Quasi-National Park (QNP), situated nearly 50 km away from the heart of Tokyo, Japan.The park's renowned highlight is the Takaosan Yakuoin, a temple founded in 744 that allows tourists to relish the amalgamation of nature, culture, and history.The site's elevation ranges between 230 to 599 meters above sea level, located at latitude 35°37' N and longitude 139°14′ E, and has several trails leading up to the mountain summit.In this study, a trail that starts at Road No. 1 and leads to the top of the mountain was used as the experimental site (Figure 1).The vegetations are natural conifers, deciduous forests, and coniferous plantations.The duration of the entrance gate to the mountain summit is approximately 90 minutes.Similar landscape assessment studies have been conducted in other trail of Takao QNP (Mizuuchi et al. 2016;Mizuuchi 2023) to identify scenic hotspots.

Geotagged visitor employed photography for on-site experiment
Visitor employed photography (VEP) focuses on people's experiences in natural environments and has been used in disciplines, such as landscape assessment, leisure studies, and tourism (Balomenou and Garrod 2016).In the VEP process, participants are asked to take photographs based on their landscape preferences, and then the results are analysed.VEP is a suitable method to assess an environment based on the actual behaviours of visitors (Hull and Stewart 1995;Oku and Fukamachi 2006;Nielsen et al. 2012;Rathmann et al. 2020).In recent years, the method was developed to geotagged VEP (GVEP).In GVEP, researchers acquire the positional data of every photograph by utilizing GNSS (global navigation satellite system) receivers and subsequently perform quantitative analyses through GIS, such as identifying scenic hotspots.GVEP serves as a valuable approach (Mizuuchi and Nakamura 2021), due to a need for GISbased research on forest recreation landscape (Domingo-Santos et al. 2011;Brown and Weber 2011;Beeco and Brown 2013).

Participants and survey protocol
One hundred fifty-three participants participated in the GVEP survey.The participants were actual visitors recruited at the entrance of the No. 1 route of Mt.Takao.The profiles of the participants are shown in Table 1.The GVEP survey was conducted in 2014: two days in spring (April), two days in summer (August and September), and one day in autumn (November); all the days had fine weather.Prior to embarking on the trail, the GVEP survey protocol was explained to the participants.They were requested to take at least 10 photographs of their favourite landscapes, objects, and places by their personal cellphones  or digital cameras.Additionally, they were given a GNSS logger, specifically the HOLUX Wireless GPS Logger M-241.After returning from the trail, all the photographs were gathered and transferred to the researcher's laptop.By matching the time codes of the GNSS logger and photographs, the photographs were converted into geotagged photographs.

Data analysis
A GIS analysis was performed to identify the scenic hotspots on the trails in the Takao QNP.Kernel density estimation is a major analysis method for identifying scenic hotspots (Chen et al. 2018).We applied network-based kernel density estimation (NKDE) using the SANET tool (Okabe et al. 2006).NKDE is an extended method for adapting kernel density estimation to linear network sites (Okabe et al. 2009), and is more suitable for such sites.The cell width was set at 10 m and bandwidth was set at 30 m.The top 40% of NKDE values expressed by the geometrical intervals were interpreted as scenic hotspots, except for two echo lift stations.The landscapes of the scenic hotspots were interpreted using a theory (Kaplan and Kaplan 1989) that spatial configuration and content-based properties (Strumse 1994;Nielsen et al. 2012;Liu et al. 2021).The authors selected one photograph to represent each scenic hotspot.

Photo-based questionnaire and respondents
Respondents were asked to complete a photo-based questionnaire that included an evaluation of the nine photographs selected in Survey 1 using 10 adjectives paired with psychological measures, including "preferable" and "naturalness" (7-point scale; see Supplement 1).The respondents were also asked to choose and rank two photographs that would attract them to visit a site; and provide a free description on the reasons for their choice.Because differences in the purpose of the evaluation could affect the evaluation (Purcell et al. 1994), a distinction was made between "preference" and "attractiveness".The respondents were 42 Japanese university students (including graduate students) in Japan and 40 Korean university students (including graduate students) in Korea.Both were educated in landscape architecture (Table 2).Koreans were selected because they had accounted for the largest number of foreign tourists visiting Japan until 2013 (Japan Travel Bureau Foundation 2014) and, still account for second largest (Japan Travel Bureau Foundation 2018) to obtain practical knowledge.In addition, Japan and Korea are neighbors and have cultural similarities.Koreans were chosen as a country with cultural commonalities because of the cultural resources present in Mt.Takao and to explore the evaluation of these resources.

Data analysis
To understand which photographs were highly appreciated, an analysis of mean (ANOM) was performed with the preferable score as a variable.The ANOM was performed for the Japanese and Korean respondents, respectively.ANOM is a graphical technique that aids in comparing mean values to ascertain whether any of them differ significantly from the population mean (Rao and Rag 2018).Next, the first photograph for "attract to go" (ATG) was counted as 2 points, the second as point.The results were then tabulated.To understand which photographs respondents were highly attracted to, an analysis of mean with transformed rank (ANOM-TR) was performed using the ATG values as variables for the Japanese and Korean participants.Then, to compare the evaluations between Japanese and Korean respondents, a Welch's t-test was performed using preferable as a variable, and a Wilcoxon rank-sum test was performed using the ATG as a variable.To compare preference and attractiveness, a Kendall's rank correlation analysis was performed and to compare preferable and naturalness, a Pearson's correlation analysis was performed.
To examine the relationship between photographs taken on-site and off-site appreciation, a Pearson's correlation analysis was performed using each photograph's NKDE score and preferable scores as variables.The first ATG photograph was counted as 2 points and the second as 1 point, and the tabulated to examine whether photographs likely to be taken on-site are more likely to attract tourists.Kendall's rank correlation analysis was performed using each photograph's NKDE value and ATG value as variables.Data analysis was performed through JMP Pro 17.0 (SAS Institute Inc., Cary, NC, U.S.A.) and the significance level for p-values was set at 0.05.

Scenic hotspots from the survey of geotagged visitor employed photography
A collection of 1,645 geotagged photographs was obtained.The NKDE analysis identified nine locations (SH1-9) as scenic hotspots (Figure 2).SH1 was a corner surrounded by forest.SH2 and SH3 were in the vicinity of an observation point.SH4 was in the vicinity of a large sacred tree.SH5 was in the vicinity of the temple gate building.SH6 was in the vicinity of the temple's main building.SH7 was in the vicinity of an observation point on the summit.SH8 was the vicinity of trees with coloured leaves taken during the autumn survey.SH9 was in the vicinity of a suspension bridge surrounded by forest.The landscapes of SH1, SH2, SH3, and SH7 were categorized as landscape types with an emphasis on spatial configurations, while SH4, SH5, SH6, SH8, and SH9 were categorized as content-based landscape types (Figure 3).

Off-site photograph appreciation
The results of the ANOM for the preferable score revealed that SH8 scored statistically higher than the mean for the Japanese population, and SH1, SH3, and SH7 scored statistically higher among the Korean participants (Figure 4).Landscape preferences differed between the Japanese and Korean participants.The Japanese participants preferred trees in autumn colours, while the Korean participants preferred natural landscape photographs.The results of the Welch's t-test revealed significantly different scores for SH4, SH5, SH6, SH8, and SH9 among the Japanese and Korean participants (Table 3).All were content-based landscape types with a large degree of human intervention.Conversely, there were no difference between Japanese and Koreans in the prospect and forest trail photos of SH1, SH2, SH3, and SH7.
The results of the ANOM-TR for ATG revealed that SH7 and SH9 scored statistically higher than the population mean among the Japanese respondents, and SH1 and SH7 scored statistically higher among the Korean respondents (Figure 5).SH7, a photograph of the view from the summit, was attractive for both Japanese and Korean participants, while only the Japanese participants found the suspension bridge attractive, and only the Korean participants found the trail with forests attractive.The results of the Welch's t-test for the ATG revealed that SH1, SH8, and SH9 showed significantly different preferences between the Japanese and Korean respondents (Table 4).Conversely, there were no difference between Japanese and Korean respondents in terms of the prospect type (SH2, SH3, and SH7).The Kendall's correlation coefficient for investigating the relationship between the preferable score and the ATG among the Japanese respondents was 0.18 (p < .001),and among the Korean respondents 0.20 (p < .001).There was a weaker correlation between preference and attractiveness.
Pearson's correlation analysis was used to determine the relationship between preferable and naturalness.The correlation coefficient for the Japanese respondents was 0.35 (p < .001),and for the Korean respondents 0.58 (p < .001).Korean respondents showed a higher correlation for the relationship between preferable and compared to the Japanese respondents.
Pearson's correlation analysis was used to determine the relationship between NKDE and preferable.There was no significant correlation in the Japanese (p = .479)and Korean (p = .063)respondents.The results of Kendall's correlation coefficient for NKDE and ATG showed no correlation in the Japanese (p = .463)and Korean (p = 0.917) respondents.

Discussion
The first objective of this study was to identify scenic hotspots based on on-site visitor experiences.The GVEP survey and GIS analysis identified nine scenic hotspots.Regarding content-based landscape types (SH4, SH5, SH6, SH8, and SH9), it could offer a clear explanation for why these locations have become scenic hotspots.Respondents were interested in trees, buildings, and suspension bridges.The landscapes of the three observation points could also be obvious.SH1 confirmed Kaplan and Kaplan's tentative theory of the corner in trails (Kaplan and Kaplan 1989;Herzog and Kirk 2005;Mizuuchi 2023), namely that corner landscapes evoke a response in visitors.These scenic hotspots are likely subjects for individual visitors' photographs, and their landscape management is considered particularly important.In addition, the views from the scenic hotspots should be considered in terms of their visual impact to create a satisfying landscape experience (Palmer 2019).
The analysis of off-site landscape appreciation revealed that the photographs that the Japanese people prefer are not similar to those that are likely to attract visitors, and were not the same as those that Korean respondents prefer to photograph as well.Korean respondents tend to prefer natural landscapes and match their preferences with the site's attractiveness.There were significant differences between the Japanese and Korean respondents, especially in terms of photographs that contribute to attracting potential visitors.The Japanese respondents found the suspension bridge attractive place to visit, but the Korean respondents did not find it attractive.The Korean respondents perceived trails surrounded by forests as attractive, whereas the Japanese respondents did not.
Although few studies have examined these specific differences, one study has explored gardens in terms of landscape preferences.Japanese gardens imitate nature and to make them look natural, while Korean gardens strictly separate areas left as natural and artificial areas (Fujii 1990).This difference in the perception of landscapes may explain why the Japanese respondents perceived photographs of suspension bridges that were designed to look natural as attractive, whereas the Korean respondents perceived the photographs of forests with little human influence other than the trail itself as attractive.There was no difference between the two countries in the evaluation of "prospect" in terms of both preferences and ATG.According to Appleton's prospect refuge theory (Appleton 1984(Appleton , 1996)), humans instinctively prefer prospect landscape.Although this theory has not been completely proven (Dosen and Ostwald 2016), "prospect" is said to be a landscape universally preferred by humans (Tveit 2009).The result of this study showed that prospect was not the landscape type that reflected cultural backgrounds; it thus partly confirmed Appleton's theory.
Many studies have encouraged national parks and destination management organisations (DMO) to use social media platforms to attract tourists (Hays et al. 2013;Pino et al. 2019;Pop et al. 2022).In fact, the Japanese National Park's administration reposts photographs posted on individual tourist accounts for promotion.However, the findings of this study suggest that all the photographs that were preferred by Japanese visitors to Takao QNP may not necessarily appeal to potential visitors.There is no correlation between the NKDE value of scenic hotspots and off-site photograph appreciation.While crowd-sourced landscape assessment studies are useful for understanding on-site landscape preferences (Heikinheimo et al. 2017;Bubalo et al. 2019;Goldberg 2019), caution should be exercised in determining whether photographs of a location are highly valued.
Regarding landscape preferences, the result of this study confirmed previous studies' findings on similarities and differences, namely that there are fewer cultural differences in natural landscapes and more differences in humaninfluenced landscapes (Kent 1993;Strumse 1996;Takayama et al. 2006).As international tourism expands, cross-cultural research findings will become increasingly important for forest landscape planning and management.

Conclusion
In the age of big data, the influence of social media on on-site landscape experiences is growing.In this study, we identified places and photographs that are highly rated as scenic on-site hotspots.We also examined their off-site appreciation in terms of preferences and attractiveness as places to visit and the sequential relationship between on-site and off-site experiences.The outcomes of hotspot analysis will contribute towards enhancing practical landscape management.While the geographical location identified as a hotspot holds a significance, it is crucial to acknowledge that the visible area from the hotspot also influences the tourist's scenic experience.Thus, visible areas from prospect type hotspots, as shown in supplement 2, indicates areas of scenic importance.Consequently, this study has yielded valuable insights for the future forest landscape management in practical terms.All the photographs that were highly evaluated on-site were not necessarily highly appreciated off-site, and there were differences between the Korean and Japanese respondents.These findings provide fundamental knowledge for user-friendly, high-quality tourism promotion.As international tourism expands, cross-cultural research on on-and off-site experiences and preferences has become increasingly important for forest landscape management and sustainable tourism.
Caution should be exercised in interpreting the results of this study since it was conducted solely in one park.
Additional research is necessary to comprehend the connection between visitor experiences on-and off-site.Moreover, since the participants were limited to landscape architecture students, it is uncertain whether the findings are indicative of a more extensive population's landscape preference.

Figure 2 .
Figure 2. The nine scenic hotspots by NKDE.NKDE values are expressed by the geometrical interval.

Figure 3 .
Figure 3.The nine photographs of each scenic hotspot.

Figure 4 .
Figure 4. ANOM results comparing the preferable scores of the nine photographs of each scenic hotspots (α = 0.05).Red and green points atop vertical black lines indicate mean scores.Horizontal black lines indicate the upper and lower decision limits (UDL and LDL, respectively).Green points within the UDL and LDL represent no statistical difference from the population rank mean, whereas red points above and below the UDL and LDL are statistically higher and lower than the population rank mean, respectively.

Figure 5 .
Figure 5. ANOM-TR results comparing the ATG scores of the nine photographs of each scenic hotspots (α = 0.05).Red and green points atop vertical black lines indicate mean scores.Horizontal black lines indicate the and lower decision limits (UDL and LDL, respectively).Green points within the UDL and LDL represent no statistical difference from the population rank mean, whereas red points above and below the UDL and LDL are statistically higher and lower than the population rank mean, respectively.

Table 1 .
Profiles of the participants engaged in the geotagged visitor employed photography survey.

Table 2 .
Profiles of the respondents engaged in the photo-based questionnaire.

Table 3 .
The result of the Welch's test(α = 0.05) to compare differences of the preferable between Japan -Korea for the photographs of each scenic hotspots.

Table 4 .
The result of the Wilcoxon sum-rank test(α = 0.05) to compare differences of the ATG between Japan -Korea for the photographs of each scenic hotspots.