Establishing Adaptation Measures Based on Landslide Potential Impact Assessment for Scenic Sites

Background and Objectives

The international community is increasingly facing various environmental changes due to the intensification of the climate crisis. According to the IPCC (2021), the global average temperature has risen by 1.09°C compared to the pre-industrial era (1850–1900). If the current trend in greenhouse gas emissions continues, the temperature will increase by 1.5°C by 2040. This climate change extends beyond simple warming, leading to increased variability in precipitation patterns and extreme weather events such as heavy rainfall, typhoons, droughts, and snowstorms with unprecedented frequency (Government Joint Report, 2022). Particularly, the Korean Peninsula is assessed to be highly vulnerable to climate change within East Asia, with climate-related disaster risks steadily increasing (Government Joint Report, 2022). The urgency of this situation cannot be overstated, and immediate action is required to mitigate these risks.

The increase in extreme weather events associated with the climate crisis also accelerates the frequency and intensity of disasters, thereby exacerbating damage to national heritage assets (CHA, 2022). National heritage sites, which hold natural and cultural value, require continuous protection and management. However, climate change has led to increased precipitation and abnormal weather conditions, accelerating physical degradation and diversifying the types and frequency of damage. According to the Cultural Heritage Administration (CHA, 2023), the number of disaster-related damage cases from 2018 to 2022 increased by approximately 116.1% compared to the previous five years (2013–2017), with storm and flood damage—particularly affecting natural heritage—rising by around 150%. Nevertheless, current adaptation measures for national heritage sites fail to adequately consider the distinct characteristics of each heritage type, and tailored strategies remain insufficient. This research aims to address this gap by proposing tailored adaptation strategies specifically designed to preserve each heritage type's unique characteristics and values.

Among the various forms of national heritage, Scenic Sites are particularly vulnerable to landslides triggered by climate change. Defined under the “Natural Heritage Act” as places of outstanding natural beauty with historical and academic value (Korea Law Information Center, 2024), Scenic Sites are landscapes where natural and cultural elements are intricately intertwined, making their scenic value a core attribute. Beyond their natural beauty, Scenic Sites are complex heritage spaces encompassing historical and cultural significance, making them essential subjects of heritage management nationally and globally. Their unique characteristics and the value they hold make them particularly significant in the context of increasing disaster risk.

Mountainous Scenic Sites are especially susceptible to changes in vegetation and topography induced by shifting rainfall patterns, resulting in a significantly higher risk of landslides. Although the risk of landslides is increasing with the rise in extreme rainfall events, Korea currently lacks customized adaptation measures that can simultaneously preserve the landscape value of Scenic Sites and provide effective disaster mitigation. Existing disaster prevention facilities and adaptation policies are mainly designed for general forested areas and do not reflect the landscape-specific characteristics of Scenic Sites. As a result, they undermine the sites’ adaptive capacity to landslides and risk damaging their scenic integrity. Therefore, moving beyond conventional disaster response strategies and establishing site-specific adaptation measures that balance landscape preservation and disaster resilience are vital.

This study aims to assess the potential impact of landslides in the Gayasan Haeinsa Scenic Site using the IPCC’s climate vulnerability assessment framework and to propose effective, landscape-conscious adaptation strategies that address the limitations of existing measures. Furthermore, field investigations are conducted to evaluate the feasibility and effectiveness of the proposed methods, ultimately contributing to the development of sustainable management approaches for Scenic Sites in the face of climate change.

Scope of the Study

The Gayasan Haeinsa Scenic Site study area is located in Hapcheon-gun, Gyeongsangnam-do, South Korea, and spans 20,952,454 square meters. The Cultural Heritage Administration designated it as a Scenic Site in 2009 (CHA, 2009). Situated within Gayasan National Park, the area includes major natural landscape features such as Duribong, Gitdaebong, Namsan Jeilbong, and the Hongryudong Valley. Among these, the Hongryudong Valley—known as the "Ten-ri Valley"—is especially renowned for its scenic beauty, often compared to the Okryucheon Valley of Mount Geumgang. The surroundings of Haeinsa Temple are home to numerous hermitages such as Wondangam, Hongjeam, and Samseonam, along with important Buddhist cultural relics from the Silla period, including the Rock-carved Buddha and the Stone Standing Buddha (designated Treasures), thereby highlighting the site's value as a historical and cultural scenic landscape (National Research Institute of Cultural Heritage, 2011).

The Gayasan Haeinsa area spans across both Gyeongsangnam-do and Gyeongsangbuk-do. The Gyeongsangbuk-do portion, in particular, is a mountainous region where forests account for approximately 70% of the total land area, making it one of the regions highly vulnerable to landslides induced by climate change. Furthermore, the area is a complex cultural landscape encompassing a high concentration of designated heritage assets. As of 2024, the boundaries of the Scenic Site include one National Treasure, eight Treasures, seven City/Province Designated Cultural Properties, and one Historic Site (Fig. 1). The clustering of such numerous heritage elements in a single space implies that, in the event of a landslide, the impact could extend beyond a mere natural disaster to a compounded threat to multiple national heritage assets.

Fig. 1

Status of the study area and national heritage within the scenic site boundary.

Method for Evaluating Potential Impact and Calculating the Potential Impact Index

To quantitatively evaluate the potential impact of landslides in the Gayasan Haeinsa Scenic Site, this study adopted the conceptual framework for climate vulnerability assessment presented in the IPCC Fourth Assessment Report (AR4). Specifically, Potential Impact was defined as the combination of Exposure and Sensitivity. IPCC (2007) defines vulnerability as “the degree to which a system is susceptible to, or unable to cope with, adverse effects of climate change, including climate variability and extremes.” It consists of exposure, sensitivity, and adaptive capacity. In subsequent reports (AR5 and AR6), the IPCC introduced a risk assessment framework, conceptualizing risk as an interaction of hazard, exposure, and vulnerability, and further expanding to include compound and cascading risks.

This framework conceptualizes exposure as the presence of climate-related physical phenomena, long-term trends, or resulting impacts that could affect a given system. Sensitivity refers to the degree to which specific areas or elements, such as land units, infrastructure, or heritage resources, are likely to be affected when exposed to such hazards. Potential impact, therefore, is defined as the inherent tendency or predisposition of social, ecological, or cultural components, such as human populations, ecosystems, heritage assets, or scenic landscape values, to suffer adverse effects from climate change (Table 1).

Table 1

Potential impact index definition

While the AR6 framework of the IPCC presents a comprehensive risk assessment approach that incorporates compound risks and socio-environmental contexts, its application requires the inclusion of adaptive capacity in the assessment of vulnerability. However, in the case of Scenic Sites in South Korea, there are currently no established disaster adaptation strategies specifically tailored to their landscape and heritage characteristics. As such, evaluating vulnerability by including adaptive capacity may lead to ambiguous or inconsistent interpretations. Therefore, this study adopts the framework of the Fourth Assessment Report (AR4), which defines potential impact as a combination of exposure and sensitivity, excluding adaptive capacity. This approach enables a more direct assessment of the inherent risk posed by landslides under climate change, with the ultimate goal of supporting the development of site-specific adaptation strategies optimized for Scenic Sites (Fig. 3).

Fig. 3

Potential impact formula.

Because the indicators used in this study have different value ranges, unnormalized data can lead to bias or over-fitting in the results. The Min-Max normalization method was applied to ensure comparability across variables, which scales each indicator to a 0–1 range, with 0 as the minimum and 1 as the maximum value (Baek, 2023). The normalized indicators were then used in the potential impact formula presented in Fig. 2 to calculate the final Potential Impact Index. Although it is possible to apply weights to each variable to reflect its relative importance, this study applied equal weights to all indicators to maintain consistency and fairness across variables.

Exposure Assessment in Scenic Sites

Exposure Assessment of National Heritage within the Scenic Site

By the definition of exposure, this study selected the vulnerability of national heritage sites located within the boundary of the scenic site—both point-based and area-based heritage properties—as one of the key indicators for exposure assessment. These heritage elements are considered susceptible to landslide hazards. The Cultural Heritage Administration (CHA) in Korea designates Historic and Cultural Environment Protection Zones to safeguard the heritage properties and their surrounding areas, which are recognized for their outstanding natural, historical, or cultural landscape value. According to the Cultural Heritage Protection and Utilization Act (2025), such zones can extend up to 500 meters from the outer boundary of the designated heritage site. This study used these Historic and Cultural Environment Protection Zones as the spatial basis for defining heritage exposure. Still, they were limited to zones within the Scenic Site boundary. Overlay analysis was then conducted for each national heritage property to quantify their exposure (Fig. 4).

Fig. 4

Setting buffer ranges for exposure analysis.

Visibility-Based Exposure Assessment

Another indicator used for exposure assessment was visibility-based exposure, which quantified the scenic value of the site using spatial analysis. To this end, key landscape resources within the Scenic Site and the viewpoints from which these resources could be observed were identified, and a viewshed analysis was conducted. Landscape resources were classified into natural and cultural categories, following the classification system proposed by the National Research Institute of Cultural Heritage (2011). Given that landslides are less likely to occur at high elevations, it was assumed that peaks would be minimally affected by such hazards. Therefore, peaks were excluded from the natural landscape category in this study. As a result, three natural landscape resources (waterfalls and valleys) and seventeen cultural landscape resources, including fifteen structures and two Buddha statues, were selected as the final targets for analysis (Table 2).

Table 2

List of selected landscape resources

The selection criteria for viewpoints were developed based on the landscape resource database published by the National Research Institute of Cultural Heritage. Four primary criteria were applied: symbolism, external visibility, visual axis, and spatial nodes. Symbolism was defined as locations with representational or cultural significance within the Scenic Site, such as main entrances or historically significant architectural structures. External visibility referred to areas within the site offering open views toward surrounding features, including hills, rivers, or green spaces. The visual axis criterion identified elevated locations that form prominent sightlines and contribute to the structural composition of the landscape. Spatial nodes were characterized as areas of frequent movement or congregation, including public facilities, entranceways, and key intersections within the site. These four criteria were used to determine suitable viewpoints, from which viewshed analyses were conducted toward the selected landscape resources to quantitatively assess the scenic value of the site in terms of visual exposure (Table 3).

Table 3

Criteria for selecting view points

Method for Assessing Exposure in Scenic Sites

Input Variables for MaxEnt Modeling

This study required a high-resolution susceptibility map to evaluate the applicability of adaptation measures in response to the potential impact of landslides in the study area. Therefore, the spatial resolution of the input variables was standardized to 10 meters by 10 meters. Due to the limited number of known landslide occurrences strictly within the boundary of the scenic site, using only those points for model training would have been insufficient. Landslide occurrence data were collected from the surrounding administrative regions—Seongju-gun, Goryeong-gun, Hapcheon-gun, Geochang-gun, and Gimcheon-si—covering five areas to address this limitation. Four hundred ninety-four landslide occurrence points recorded over the past 10 years (2011–2021) were used as presence data for the MaxEnt model (Fig. 5).

Fig. 5

Presence data for MaxEnt modeling.

Previous studies on landslide susceptibility modeling using MaxEnt informed the initial selection of environmental variables. Cha et al. (2023) limited climatic variables to rainfall-related factors, reflecting the characteristics of landslide occurrences in South Korea, while Lee et al. (2017) utilized topographic and soil variables, and Kim et al. (2013) incorporated vegetation-related variables. In light of the above, seven ecological variables were ultimately used, encompassing climatic, topographic, geological, soil, and vegetation factors. These variables were selected based on their theoretical relevance to landslide occurrence in mountainous cultural landscapes such as Scenic Sites, the availability of high-resolution spatial data, and the need to represent diverse environmental domains.

Climatic variables were derived from gridded datasets in MK-Prism version 2.1 (2000–2019) provided by the Korea Meteorological Administration and were downscaled to match the target resolution. Topographic data were obtained from a 1:5,000-scale digital elevation model (DEM) provided by the National Geographic Information Institute. Geological and soil information was rasterized from detailed soil maps produced by the Rural Development Administration’s “Soil Map” database. Vegetation variables were sourced from national forest type maps provided by the Korea Forest Service and converted into raster format (Table 4).

Table 4

Input variables and their characteristics used in MaxEnt modeling

Criteria for Selecting Adaptation Measures Considering Potential Landslide Impact in Scenic Sites

Based on the landslide vulnerability assessment results, this study compared the Potential Impact Index of each type of landscape resource within the scenic site. Resources with relatively higher impact scores were identified as Key Management Landscape Resources.

These selected resources were then used as pilot targets for evaluating the applicability of adaptation measures that reflect landslide risks specific to the scenic site. Field investigations were conducted in parallel, and optimal adaptation strategies were derived by considering the local environmental characteristics, including topography, soil conditions, and vegetation. The goal was to ensure that proposed measures would not only be effective in mitigating landslide risks but also preserve the scenic and cultural values of the site.

According to the Forest Protection Act, South Korea currently mandates the establishment of a National Long-Term Landslide Prevention Plan every five years. The Korea Forest Service also prepares an annual National Comprehensive Landslide Prevention Plan, which serves as the basis for managing landslide hazards nationwide. The currently implemented adaptation strategies are primarily divided into structural and non-structural approaches. Structural measures mainly include check dams and slope stabilization techniques, while non-structural measures focus on long-term monitoring, community education, and emergency evacuation training.

However, one of the primary concerns with conventional structural approaches—particularly check dams—is that these structures are often constructed with concrete, which can significantly disrupt the natural scenery. As a result, newer approaches are being proposed that reduce visual impact on the landscape, such as vegetated soil bag methods and vegetation-based slope stabilization techniques. These are more compatible with the preservation needs of Scenic Sites.

Accordingly, this study reviewed both existing adaptation strategies and those proposed in domestic and international literatures, and assessed their applicability to the Key Management Landscape Resources. Considering factors such as heritage density and visitor movement patterns, the study designated Key Management Areas based on the boundaries of Historic and Cultural Environment Protection Zones in Gyeongsangnam-do. These areas were then used to inform proposals for the placement of future disaster mitigation infrastructure (Fig. 6).

Fig. 6

Criteria for setting up key management area.

Landslide Sensitivity Analysis Results

Based on the selected variables, the MaxEnt model was executed to predict landslide sensitivity spatially within the study area. The resulting map illustrates the probability of landslide occurrence, ranging from 0.00 to 0.86 (Fig. 7). Higher values indicate areas with greater susceptibility to landslides.

Fig. 7

Landslide Susceptibility map of the scenic site.

As shown in the sensitivity map, the northwestern and southeastern regions of the study site exhibit high sensitivity to landslides. In contrast, areas with lower elevation or gentle slopes demonstrate relatively low sensitivity.

To evaluate the predictive performance of the model, a Receiver Operating Characteristic (ROC) analysis was conducted, and the Area Under the Curve (AUC) value was calculated (Fig. 8). The study showed that the model achieved an AUC value of 0.765, indicating a satisfactory level of predictive performance. An AUC value greater than 0.5 suggests that the model performs better than random prediction, and it is generally accepted that a value above 0.7 indicates good explanatory power (Swets JA, 1988).

Fig. 8

Model accuracy assessment using the ROC curve.

Nevertheless, this value does not represent the highest possible level of explanatory power, which may be attributed to reduced model performance resulting from the downscaling of coarse-resolution climate data. The spatial uncertainty introduced during this process may have limited the accuracy of climatic input variables, thereby affecting overall model predictability.

Response curves were analyzed for each input variable (Fig. 9). Regarding annual precipitation, data from the Korea Meteorological Administration indicate that the national average is approximately 1,320.3 mm. In the response curve of this study, the probability of landslide occurrence increased sharply when precipitation exceeded 1,300 mm, suggesting that areas with higher-than-average rainfall may be more susceptible to landslides.

Fig. 9

Susceptibility map of the scenic site.

For maximum daily precipitation, the highest probability of landslide occurrence appeared around 120 mm, after which the probability decreased with increasing precipitation, showing a nonlinear pattern. This may be attributed to the nature of the dataset, which reflects the average annual extreme daily rainfall between 2000 and 2019, potentially underrepresenting the impact of high-intensity rainfall during landslide events.

Regarding elevation, the probability of landslide occurrence was relatively high below 500 m and decreased above 700 m. This trend aligns with the findings of Kim et al. (2000), who reported that most landslides in Korea occur on slopes at elevations below 500 m.

Regarding parent rock material, landslide occurrence was highest in areas underlain by granite (code 1) and metamorphic rock (code 5), with granite showing the most significant susceptibility. This result is consistent with the findings of Kim et al. (1998), which identified granite as significantly more vulnerable to landslides than metamorphic rock.

For forest type, all categories except non-forested land showed similar patterns. Although Jung (2010) reported that mixed forests accounted for 51.2% of landslides in Gyeongsangbuk-do, the relatively uniform distribution of forest types in the Haeinsa area may have contributed to the lack of marked differences. This also suggests that topographical and rainfall-related factors may have substantially influenced landslide occurrence more than forest type.

About forest age class, the probability of landslide occurrence was relatively higher at specific age intervals, though no consistent trend was observed, and variation between age classes was prominent.

Visibility Exposure Analysis of the Scenic Site

Based on the selected viewpoints, a cumulative viewshed analysis was conducted to evaluate the visibility exposure of the study area (Fig. 10). Visibility exposure was quantified by assessing how frequently each location within the scenic site was visible from multiple viewpoints. This allowed for a spatially explicit and quantitative understanding of the site's exposure patterns.

Fig. 10

Results of visibility frequency analysis from each viewpoint.

The analysis revealed that certain areas, particularly around the southwestern Jeilbong Peak and the vicinity of Haeinsa Temple, exhibited the highest levels of visibility. These areas were repeatedly visible from several different viewpoints, suggesting their significance in the scenic composition of the site. In contrast, areas with lower visibility were likely obstructed by surrounding terrain or vegetation, resulting in reduced visual exposure.

A spatial density map of national heritage sites was generated, and an overlay analysis was conducted to identify areas with high concentrations of heritage elements (Fig. 11). This analysis aimed to quantify the clustering of heritage sites in specific areas, thereby identifying regions with a higher potential for concentrated damage in the event of a landslide.

Fig. 11

Results of overlay analysis on the exposure of heritage elements within the scenic site.

The results indicated a notable clustering of national heritage sites near Haeinsa Temple. This suggests that the area may be particularly vulnerable to extensive damage should a natural disaster, such as a landslide, occur. In areas with high heritage density, multiple assets may be simultaneously affected, highlighting the need for targeted protection and management strategies.

Results of Potential Impact Assessment and Identification of High-Risk Landscape Resources

A final landslide potential impact map was created by summing the standardized exposure and sensitivity values. The results showed that higher potential impact index values corresponded to areas with increased susceptibility to landslide impacts within the scenic site (Fig. 12). The core management zone, located at the center of the study area, exhibited a high potential impact due to the concentration of multiple cultural heritage assets. This indicates a greater likelihood of direct exposure to landslide hazards. In contrast, the peripheral areas outside the core zone generally showed lower levels of potential impact.

Fig. 12

Landslide potential impact map of the scenic site.

A comparative analysis of the distribution of nationally designated and province/city-designated heritage sites between the core management zone and its outer areas revealed that the core zone includes ten nationally designated heritage items—two National Treasures, six Treasures, and two Historic Sites—as well as six City/Province Designated Cultural Heritage sites. In contrast, the outer zone contains relatively fewer heritage resources, including only three province/city-designated sites (two Treasures and one Tangible Cultural Property). This disparity in distribution suggests that the heritage assets within the core management zone are significantly more vulnerable to landslide impacts.

Each landscape resource's average potential impact index was calculated based on the raster cells within a 500-meter radius. Using these values, the resources with the highest possible impact were identified by type and designated as Key Management Landscape Resources. Among the built heritage resources, Gumseonam exhibited the highest potential impact with a value of 1.67, followed by Hongyeam at 1.61 and Yongtap Seonwon at 1.59. In the valley category, Hongryudong Valley Section 3 showed the highest potential impact with a value of 1.22. In contrast, waterfall resources appeared to be less affected. Nakhwadam recorded a value of 0.53, while Yongmun Falls showed 1.50. As for the Buddha statue resources, which are located in forested highland areas outside the core scenic zone, they exhibited relatively low potential impacts—the Rock-carved Standing Buddha and the Stone Standing Buddha had values of 0.61 and 0.45, respectively (Table 5).

Table 5

List of landscape resources’s potential impact index

Accordingly, Gumseonam, Hongryudong Valley Section 3, the Rock-carved Standing Buddha, and Nakhwadam—which exhibited the highest potential impact values within each landscape resource category and were located within or adjacent to the key management area—were initially identified as key management landscape resources. However, since the Rock-carved Standing Buddha and Nakhwadam are situated outside the key management area and their potential impact values were relatively low, this study focused on Gumseonam and Hongryudong Valley Section 3, both of which are situated within the key management area, for the development of adaptive measures (Fig. 13).

Fig. 13

Status of key management landscape resources(A: Geumseonam, B: Hongryudong Valley, C: Nakhwadam, D: Rock-carved standing buddha).

Adaptive Measures Considering the Potential Impact of Landslides on the Scenic Site

Selection of Applicable Adaptive Measures

To assess the applicability of adaptive measures in response to landslide impacts on the selected key management landscape resources, this study compiled a list of adaptation strategies currently implemented by the Korea Forest Service. In addition, adaptive measures that help preserve the scenic integrity of the site—such as vegetation mat techniques—were considered and included in the list (Table 6).

Table 6

Landslide prevention measures

Applicability Assessment of Adaptive Measures for Key Management Landscape Resources

Field investigations were conducted for each key management landscape resource to assess the applicability of adaptive measures. In the case of Hongryudong Valley Section 3, structural adaptive measures were deemed necessary, particularly for the left side of the valley and adjacent areas where the potential impact index was relatively high compared to surrounding regions (Fig. 14). The site was found to be a natural stream terrain characterized by steep slopes and exposed rock surfaces. Therefore, installing cement-based stabilization nets and designing retaining structures were identified as necessary interventions. In addition, due to the site’s topographic characteristics that cause concentrated surface runoff during rainfall, the installation of check dams was suggested to control water flow. This should be complemented by a drainage system using culverts to prevent slope instability caused by localized water accumulation.

Fig. 14

View direction (Left) and Site view (Right) of Geumseonam.

In the area surrounding Gumseonam, a high potential impact was identified on the steep, visually obscured slope to the left of the hermitage. Field surveys revealed significant vegetation degradation and a risk of soil erosion on this slope. Considering these conditions, the application of a vegetation-based stabilization technique, such as the vegetated soil bag method, was deemed appropriate to ensure both slope stability and the preservation of the forest landscape (Fig. 15). Additionally, it is recommended to install landslide sensors on the upper parts of the high-risk slope to establish a real-time monitoring system. Integrating a weather monitoring system is also considered reasonable for a more comprehensive risk assessment.

Fig. 15

View direction (Left) and site view (Right) of Hongryudong valley.