Introduction
As urbanization progresses, forest landscapes become increasingly fragmented, resulting in reductions in both the extent and quality of wildlife habitats (Haddad et al., 2015). Habitat loss and fragmentation make it progressively more difficult for wildlife to remain within their original natural environments, prompting and expansion of their activity ranges into areas adjacent to human-dominated landscapes (Soulsbury and White, 2015). In particular, urban green spaces, such as mountains an parks, as well as riparian zones, are intensively used by the public, thereby increasing the likelihood of human-wildlife encounters. These interactions can give rise to a range of socio-economic challenges, including threats to human safety, damage to property and agricultural crops, and elevated levels of public concern and anxiety (Schell et al., 2021; Galley et al., 2024; Pant et al., 2023)
In South Korea, human-wildlife interactions have likewise increased in urban and peri-urban areas, with wild boars accounting for the majority of reported damage cases (Kang, 2025). Wild boars (Sus scrofa) are highly adaptable species, capable of flexibly adjusting their behavior, movement patterns, and resource use in response to environmental change and habitat disturbance. Studies conducted in European urban contexts have shown that wild boars exploit fragmented green spaces, accessible food resources, and movement corridors to establish and expand their populations within and around urban areas (Stillfried et al., 2017; Podgórski et al., 2013; Castillo-Contreras et al., 2018).
In Seoul, approximately 1,470 reports of wild boar sightings were recorded over the three-year period from 2021 to 2023, raising concerns about the safety of residents living in forested urban areas and neighborhoods adjacent to woodlands. Accordingly, the presence of wild boars in urban environments has emerged as a persistent challenge in urban wildlife management (Seoul Metropolitan Government, 2024; Seoul Open Data Plaza, 2025). However, short-term measures that primarily focus on capture and removal have proven insufficient for achieving sustainable mitigation. There is therefore a growing need for a mid-to-long-term management framework that reduces the risk of human-wildlife conflict while promoting safe coexistence between humans and wildlife (González-Crespo et al., 2018; Escobar-González et al., 2024; Pooley et al., 2021; Conejero et al., 2024).
To address this challenge, a quantitative understanding of the environmental conditions influencing wild boar occurrence and habitat use is essential (Caruso et al., 2018; Podgórski et al., 2013). Although habitat analyses based on field signs of wild boars have been conducted in forested regions in South Korea (Rho et al., 2015; Lee et al., 2022), such evidence typically reflects dynamic behaviors such as movement and foraging. This makes it difficult to determine the precise temporal context of occurrence, thereby limiting its utility for accurate habitat characterization (MacKenzie et al., 2002; Higashide et al., 2021; Massei et al., 2018). Moreover, previous studies have largely relied on two-dimensional environmental datasets, such as land cover and forest type maps, which do not adequately capture the three-dimensional structural attributes of forests that are critical for wild boar habitat selection—such as canopy density and understory vegetation development (Acebes et al., 2021; Elliott et al., 2025; Valderrama-Zafra et al., 2022). Wild boars are known to prefer understory vegetation with high canopy closure for resting and to utilize multilayered vegetation structures that provide visual concealment during movement (Ciach et al., 2022; Fradin et al., 2023; Hidalgo-Toledo et al., 2025). Therefore, integrating direct observation data on wild boar occurrence with high-resolution three-dimensional forest structural data may help overcome the limitations of previous approaches (Burns et al., 2025).
The objective of this study is to analyze the spatial patterns of wild boar occurrence probability in relation to environmental factors and to identify areas within urban parks that present a relatively high risk of wild boar presence. The findings are expected to inform the development of spatially targeted management frameworks that enhance public safety while promoting coexistence between humans and wild boars in urban ecosystems.
Research Methods
Study Site and Target Species
This study was conducted in Ansan Urban Nature Park, located in Seodaemun-gu, Seoul, Republic of Korea (total area: 1.8 km2; maximum elevation: 295.9 m). The study site is an urban forest park characterized by the dominance of Korean red pine (Pinus densiflora Siebold and Zucc.) and Korean pine (Pinus koraiensis Siebold and Zucc.) along the ridgelines. The vegetation exhibits features of a mixed forest, with oak communities—including Quercus serrata Murray, Quercus acutissima Carruth., and Quercus × mccormickii Carruth.—distributed throughout the area. In addition, large-scale plantations of metasequoia (Metasequoia glyptostroboides Hu and W. C. Cheng) and Korean pine have been established around the Ansan Jarak-gil trail in the northwestern section of the park (Moon and Song, 2015). Ansan Urban Nature Park is adjacent to urban residential areas and is intersected by numerous hiking trails, resulting in high levels of use by local residents and visitors. However, a recent increase in the occurrence of wild boars in areas surrounding the park has raised concerns regarding public safety (Cho, 2024; Pyo, 2025).
The target species of this study, the wild boar (Sus scrofa), is a predominantly nocturnal mammal that prefers habitats with dense vegetation and high canopy closure for predator avoidance and resting. It also exhibits ecological traits associated with the use of multi-layered vegetation structures, which provide visual cover when moving around (Ciach et al., 2022; Fradin et al., 2023; Górecki et al., 2009).
Data Acquisition and Preprocessing
Traditional field surveys are susceptible to over- or underestimation of species occurrence due to variability in trace detectability influenced by environmental conditions, as well as sampling bias arising from surveyor accessibility (Dröge et al., 2020; Heske et al., 2011). To address these limitations, this study collected wild boar occurrence data using unmanned sensor cameras (camera traps) and unmanned aerial vehicles (drones).
A total of 84 camera traps (Spec Ops Elite HP4; Browning, USA) were deployed from February to April 2024 to record wild boar presence. Preliminary field inspections, combined with drone-assisted surveys, were conducted to identify and verify potential wild boar habitats. To minimize spatial autocorrelation while reflecting habitat characteristics, the study area was divided into 84 square grids at 150 m intervals, with one camera trap installed in each grid (Fig. 1). Camera traps were operated in both video and continuous photo modes. To reduce duplicate records resulting from consecutive captures of the same individual, only independent detection events were retained by applying a 30-minute interval threshold. During the survey period, wild boars were detected at 31 locations. Among these, 14 locations with only a single detection event were interpreted as transient movement paths and were therefore excluded from the occurrence dataset (Kays et al., 2021). Ultimately, 17 locations with two or more repeated detections were considered representative of actual habitat use or frequently utilized areas, and were used as occurrence data for species distribution modeling (Gao et al., 2024).
Drone-based occurrence data were collected between December 2023 and May 2024, with surveys conducted primarily during dawn and dusk, when wild boars exhibit peak activity. Five drones equipped with thermal imaging sensors (Matrice 300 RTK, Matrice 30T, Mavic 3T; DJI, China) were simultaneously deployed to survey the entire study area. All drones were equipped with 640 × 512 pixel thermal sensors, capable of detecting ground objects as small as 50 cm, which is sufficient to identify both juvenile and adult wild boars. Drone flights were standardized at an altitude of 100 m above ground level (AGL) and a flight speed of 3 m/s. In the thermal imagery, repeated detections within a 10 m radius during the same survey date and session were treated as identical events and merged to remove duplicate coordinates, resulting in 31 validated occurrence points.
LiDAR point cloud data for the entire study area were collected in April 2024 using a Matrice 300 RTK (DJI, China) equipped with a Zenmuse L1 sensor. To improve positional accuracy, real-time kinematic (RTK) correction was applied using a D-RTK 2 base station. Mapping flights were conducted in terrain-following mode using the DJI Pilot 2 application, maintaining an altitude of 130 m AGL and a flight speed of 3 m/s. This approach produced high-density point cloud data with an average density of 816 pts/m2. The raw data were exported in LAS format using DJI Terra and subsequently processed in LiDAR360 (GreenValley International) for noise removal and ground point classification. Based on the processed data, a digital elevation model (DEM) and a canopy height model (CHM) were generated. From the DEM, topographic variables— including slope, northness (cosine-transformed aspect; cos(aspect)), and the topographic position index (TPI)— were derived using QGIS. To characterize forest structure, canopy cover was calculated to represent upper-layer density, while understory density was quantified by segmenting the vertical forest structure into height intervals and extracting vegetation density near the ground layer.
Model and Environmental Variable Selection
Species distribution models (SDMs) are widely used to quantify habitat-use characteristics and predict potential habitat suitability based on relationships between species occurrence and environmental variables (Elith and Leathwick, 2009; Guisan and Thuiller, 2005). In the case of wild boars, which frequently occur at urban-forest interfaces, spatial use and occurrence patterns are strongly influenced by complex interactions among environmental factors, including topography, land cover, and vegetation structure. Accordingly, SDMs provide an effective framework for identifying potential habitats and informing management priorities (Lee et al., 2018; Faustini et al., 2025; Yang et al., 2024). In this study, the MaxEnt (Maximum Entropy) model was employed, which requires only presence data and effectively captures nonlinear relationships (Merow et al., 2013; Phillips et al., 2006). MaxEnt has been extensively applied in wildlife habitat modeling because it demonstrates robust predictive performance even with limited occurrence records. Furthermore, it effectively integrates high-dimensional environmental variables, including indicators of vertical vegetation structure, to precisely model habitat suitability (Tattoni et al., 2012; Elliott et al., 2025; Ficetola et al., 2014; Lawrence et al., 2024).
A total of 48 wild boar occurrence points were collected from drone surveys and camera trap data and used as input for the model. Environmental predictors included four topographic variables: elevation (DEM), slope, northness, and topographic position index (TPI). In addition, vegetation variables were included, comprising forest type (Korea Forest Service, 2024) and two vertical structure metrics derived from drone-based LiDAR data, namely the canopy height model (CHM) and understory density. During the preliminary variable selection process, canopy cover was initially considered; however, it exhibited a high Spearman correlation coefficient (ρ = 0.715) with CHM. Although the variance inflation factor (VIF = 2.42) did not indicate severe multicollinearity, CHM was retained as the primary variable because it more directly represents forest vertical structure, which is central to this study. Consequently, canopy cover was excluded from the final analysis. Given the ratio of occurrence points to predictor variables, a total of seven environmental variables were ultimately selected for model construction (Table 1). All variables were resampled to a spatial resolution of 15 m and standardized to a common coordinate system (EPSG:5179) prior to inclusion in the SDM.
Model parameters were set with a regularization multiplier of 1.0, the feature class was set to “Auto Features,” and model performance was evaluated using five-fold cross-validation. Model performance was assessed using the area under the curve (AUC) in the receiver operating characteristic (ROC) curve (Merow et al., 2013). AUC values range from 0.5 (no better than random) to 1.0 (perfect discrimination), and are commonly interpreted as follows: 0.7–0.79 indicates “fair,” 0.8–0.89 indicates “good,” and values above 0.9 indicate “excellent” performance (Kim et al., 2024). Finally, a habitat suitability map was generated, and variable contributions along with response curves were analyzed to quantitatively evaluate the key environmental factors influencing the probability of wild boar occurrence.
Results and Discussion
Predicting the Potential Occurrence of Wild Boars in Ansan Urban Nature Park
The species distribution model (SDM) developed in this study spatially represented the potential occurrence of wild boars as continuous probability values ranging from 0 to 1. The model achieved an area under the receiver operating characteristic curve (AUC) of 0.758, indicating an acceptable level of predictive performance (Kim et al., 2024). This result suggests that the model effectively discriminates between occurrence and background points compared to random prediction (AUC = 0.5). Such performance is consistent with previously reported results for wild boar habitat and management-related distribution models in South Korea. In particular, AUC values in the range of approximately 0.7 have been commonly observed in studies employing single-algorithm approaches or ensemble modeling techniques (Lee et al., 2022; Kim et al., 2023; Ko et al., 2023).
Environmental Factors Influencing the Potential Occurrence of Wild Boars
An analysis of environmental variables influencing the potential occurrence of wild boars indicated that understory density (31.0%) had the greatest contribution, followed by slope (21.0%), elevation (20.0%), and aspect (15.7%) (Table 2). Habitat suitability remained low at minimal levels of understory density but increased sharply within the 0.08–0.3 range (Fig. 2a). This pattern suggests that once understory vegetation exceeds a certain threshold, favorable habitat conditions for wild boars are rapidly established. Suitability increased gradually with slope between 0° and 20°, followed by a pronounced rise in the 25–40° range (Fig. 2b). In terms of elevation, habitat suitability peaked at mid-altitudes around 160–180 m (Fig. 2c). The aspect variable exhibited relatively high suitability at both ends of the gradient (south-facing slopes: −1; north-facing slopes: +1; Fig. 2d).
The identification of understory density and slope as primary factors suggests that, in highly disturbed environments such as urban parks, wild boars may utilize vegetation structure and topography to minimize exposure and reduce the risk of human encounters (Stillfried et al., 2017; Lee et al., 2022; Ciach et al., 2022).
Understory density directly reflects the availability of concealment; wild boars are known to rest, wait, and move while minimizing visual detection in areas where understory vegetation is developed to a certain level or higher, and they tend to select breeding sites characterized by high understory vegetation complexity and richness (Fradin et al., 2023; Stoakley et al., 2025). Slope and elevation should be interpreted not merely as topographic preferences but in relation to human accessibility and intensity of use in urban parks (Wielgus et al., 2024; Zheng et al., 2023). In this context, the importance of slope as a key variable is consistent with the findings of Lee et al. (2022) in Bukhansan National Park, where wild boar traces were concentrated on slopes of 10–20°, similar to the present study. In contrast, studies conducted in general forest environments in South Korea have reported that wild boars prefer higher elevations and gentle ridges, suggesting that their fine-scale topographic preferences vary depending on terrain conditions and levels of human disturbance (Kim et al., 2019). Accordingly, in highly disturbed environments such as urban parks, the higher suitability observed at steeper slopes and mid-elevations may reflect the boars’ behavioral strategy to avoid human presence (Stillfried et al., 2017).
The polarized pattern observed for aspect likely reflects the topographic structure of the study area. The main ridge of Ansan Urban Nature Park extends along a northwest-southeast axis, dividing slopes into north- and south-facing aspects with potentially distinct vegetation structures. In particular, the relatively well-developed understory vegetation on the north-facing slopes may have reinforce the suitability pattern associated with understory density, suggesting that vegetation structure plays a key role in determining habitat suitability for wild boars (Rho et al., 2015; Kim et al., 2019).
The habitat suitability map showed that areas of relatively high suitability formed continuous bands along ridges and slopes, suggesting that wild boar spatial use and movement are structured along specific topographic axes within the park (Fig. 3). Notably, ridge sections predicted to have high suitability feature elevated deck pathways used by visitors heading toward the summit. In these areas, people tend to move quickly rather than stop to appreciate the surrounding vegetation. The elevated deck structure installed on the steep terrain effectively separates human activity from the ground surface.
In contrast, the northwestern section exhibited generally low suitability, which seems to reflect site characteristics where visitor facilities—such as the Metasequoia Trail, Korean Pine Forest, Forest Stage, and Red Clay Trail—are concentrated. This pattern is consistent with previous findings that increased human activity reduces wild boar habitat use in urban-edge parks (Ikeda et al., 2019). Especially in areas with concentrated human activity, vegetation management practices—such as understory clearance to improve trail accessibility and ensure visibility for safety— further reduce available cover, thereby limiting habitat use by wild boars (Ikeda et al., 2019; Tamura et al., 2024). Consequently, the key environmental variables influencing wild boar habitat selection appear to interact in ways that conflict with their ecological requirements in these areas, resulting in the extensive distribution of low-suitability patches in the northwestern portion of the park.
Conclusion
This study integrated high-resolution three-dimensional data derived from drone-based LiDAR with a species distribution model to identify environmental factors influencing wild boar habitat use in urban parks and to predict areas with high occurrence probability. First, understory density was identified as the most influential factor determining wild boar occurrence, indicating that the availability of cover plays a critical role in habitat use within urban park environments. Second, slope and elevation were associated not with intrinsic topographic preferences but with human accessibility and intensity of use within parks, with a higher likelihood of wild boar occurrence observed in steep, mid-elevation zones. Third, aspect reflected the topographic structure of the study area, with relatively higher occurrence probabilities on both north-and south-facing slopes. This pattern can be interpreted as the combined effect of local microclimatic conditions and understory vegetation development. Finally, the habitat suitability map revealed a continuous, band-like distribution of high-probability areas along ridges and slopes, suggesting that wild boars utilize habitats spatially structured along connected terrain features. These findings highlight the need to shift management and response strategies from spot-based, reactive measures toward line-based, spatial planning approaches.
The predictive outcomes provide a valuable foundation for spatially targeted management strategies aimed at promoting coexistence between humans and wild boars in urban parks. Specifically, areas with high predicted occurrence probabilities could be prioritized for preemptive management strategies, such as the installation of ecological corridors and barrier fences or designation as focused patrol zones. In the short term, measures such as installing informational signage or adjusting park usage hours in these areas may help reduce direct encounters between humans and wild boars. However, because the predictive model was developed using data from a single park, the study has limitations in terms of spatiotemporal generalizability. Moreover, the model did not fully incorporate variables directly related to human activity, such as trail use intensity, infrastructure, lighting, or artificial food sources. Future research should enhance the reliability and practical applicability of the model through external validation using independent datasets from different parks and time periods, as well as through spatiotemporal cross-validation that explicitly incorporates anthropogenic factors, including lighting, trail networks, and artificial feeding resources.
PDF Links
PubReader
ePub Link
Full text via DOI
Download Citation
