Using MaxEnt to Predict <italic>Bombus Ignitus</italic> Distribution in Urban Environments

Chae Hyeon Kang; Ye Inn Kim; Yun Eui Choi; Won Seok Jang

doi:10.11628/ksppe.2025.28.3.321

J. People Plants Environ > Volume 28(3); 2025 > Article

Kang, Kim, Choi, and Jang: Using MaxEnt to Predict Bombus Ignitus Distribution in Urban Environments

Research Paper

Ecological Landscaping

Journal of People Plants Environment 2025;28(3):321-337.

Published online: June 30, 2025

DOI: https://doi.org/10.11628/ksppe.2025.28.3.321

Using MaxEnt to Predict Bombus Ignitus Distribution in Urban Environments

Chae Hyeon Kang¹, Ye Inn Kim², Yun Eui Choi³, Won Seok Jang³^*

¹Undergradute Student, Department of Ecological Landscape Architecture Design, Kangwon National University, Chuncheon 24341, Republic of Korea

²Ph.D. Student, Department of Landscape Architecture, Kangwon National University, Chuncheon 24341, Republic of Korea

³Assistant Professor, Department of Ecological Landscape Architecture Design, Kangwon National University, Chuncheon 24341, Republic of Korea

^*Corresponding author: Won Seok Jang, wsjang@kangwon.ac.kr, https://orcid.org/0009-0000-1331-2917

^{First author} Chae Hyeon Kang, hailmytrbl@gmail.com, https://orcid.org/0009-0005-1384-5912

Received January 16, 2025, Revised February 15, 2025, Accepted April 24, 2025,

This is a Peer-Reviewed Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

ABSTRACT

Background and objective: Urbanization significantly affects pollinators, which are crucial for maintaining biodiversity and supporting ecosystem services. This study aims to predict the potential distribution of Bombus ignitus in Seoul using the Maximum Entropy Algorithm (MaxEnt) model, focusing on the influence of factors of the urban and natural environment on habitat suitability.
Methods: Using 21 occurrence records from the Global Biodiversity Information Facility (GBIF) and 10 environmental variables, two MaxEnt scenarios were developed: one including urban variables, and the other excluding them. Model performance was assessed using the Area Under the Curve (AUC), and habitat suitability was classified into four levels.
Results: Both scenarios demonstrated high accuracy, with minor differences. The key factors influencing habitat suitability were forest proximity and temperature, while urban environmental variables had limited impact. Farmlands and cultivated grasslands were identified as important habitats, providing key floral resources and nesting sites. Optimal habitats were primarily located in forested, green, and agricultural areas.
Conclusion: Natural factors, particularly forest proximity, strongly influence Bombus ignitus habitats. Urban planning should prioritize green spaces to enhance pollinator conservation and biodiversity, ensuring ecological balance in urban environments.

Key words: ecosystem services, habitat suitability, pollinator conservation, urbanization, MaxEnt

Introduction

Approximately 65% of plant species on Earth rely on insects for pollination, and more than half of angiosperms require pollinators for reproduction (Barth, 1985). Among these, bees play a critical role as pollinators, facilitating the reproduction of nearly 170,000 out of more than 200,000 species of flowering plants (Nabhan and Buchmann, 1997; Jha et al., 2005). Bee-mediated pollination supports essential ecological functions closely linked to human well-being, including ecosystem services such as agricultural productivity and the maintenance of natural landscapes (Morandin and Winston, 2005; Martins et al., 2015).

Urbanization is increasing rapidly worldwide, leading to the expansion of urban areas. This growth significantly alters the structure and function of habitats for essential pollinators such as bees. This alteration has resulted in the emergence of various issues, including reduced plant diversity, increased air pollution, and expanded coverage of impervious surfaces. These changes negatively affect pollinators' behavior, reproductive success, and colony stability (McKinney, 2008). It has been shown that transformations in urban environments directly influence pollinator habitats and distribution, ultimately reducing pollination efficiency (Goulson, 2003). Moreover, the decline in pollinators disrupts plant-animal interactions within urban ecosystems, and threatens biodiversity balance (Potts et al., 2010). Yet despite the critical role played by pollinators, research on pollinator habitats in urban environments remains insufficient. To increase urban biodiversity and sustain ecosystem services, it is essential to conduct in-depth, quantitative research on pollinator activity and habitat conditions in urban areas (Matteson and Langellotto, 2010).

Approximately 6% of plant species are classified as poricidal plants, which possess small structural pores in their anthers, making them inaccessible to general pollinators (Velthuis and Doorn, 2006). These plants have co-evolved with specific pollinators, such as Bombus ignitus, which use buzz pollination to release pollen (Vallejo-Marin, 2019). According to Lee and Dumouhel (1999), approximately 20 species of bumblebees inhabit South Korea. However, research on Bombus ignitus, a native species, remains limited (Kim, 1998, 2005; Yoon et al., 2018). The existing research has primarily focused on topics such as oviposition stimulation and colony development (Lee et al., 2005; Park et al., 2004; Yoon and Kim, 2002; Yoon et al., 2003), as well as diseases (Choi et al., 2009; Choi et al., 2010; Min et al., 2017). Nevertheless, there is a scarcity of studies exploring Bombus ignitus' resource preferences, habitats, and responses to urban environmental changes.

To address this research gap, it is crucial to apply various analytical methods, such as species distribution models, to better understand the habitat and resource preferences of Bombus ignitus in urban environments. Species distribution models predict potential species distribution by integrating environmental data with species occurrence records. These models are generally categorized into two types: those using both presence and absence data, and those relying solely on presence data (Franklin, 2010). However, the absence of a species does not necessarily indicate habitat unsuitability, which limits the accuracy of such analyses (Phillips et al., 2006; Franklin, 2010). To overcome this limitation, models using only presence data, such as Environmental Niche Factor Analysis (ENFA), Genetic Algorithm for Rule Set Prediction (GARP), Surface Range Envelope (SRE), and Maximum Entropy Algorithm (MaxEnt), are widely applied. Of these, the MaxEnt model is particularly effective for predicting habitat suitability based solely on occurrence records, as it minimizes overfitting and offers high predictive accuracy (Phillips et al., 2006; Elith et al., 2011; Song and Kim, 2012).

An earlier study predicting the potential habitat of bumblebees using species distribution models was conducted by Kim (2021). However, this research had limitations, as it did not reflect recent changes in urban environments. In rapidly evolving urban landscapes, it is crucial to use updated data to analyze pollinators' habitat preferences and resource utilization patterns while thoroughly evaluating the impacts of urbanization. Such analyses can help identify the habitat features, resources, and landscape characteristics that pollinators prefer, providing deeper insight into their adaptive strategies within urban ecosystems. Ultimately, these insights can contribute significantly to promoting the sustainability of urban ecosystems.

This study aims to predict the potential distribution of Bombus ignitus in urban environments and analyze how urban and natural environmental factors influence habitat suitability and resource preferences. By identifying factors influencing habitat suitability, we aim to provide insights for pollinator conservation. To achieve these goals, we used occurrence data for Bombus ignitus in addition to variables related to food resources, topography, climate, and urban factors to assess influences on habitat suitability. The findings from this study are expected to contribute to the development of conservation strategies for pollinators, including sustainable urban ecosystem management plans.

Research Methods

Study area and focal species

The study area, Seoul, South Korea (605.2 km²), is a highly urbanized city with diverse green spaces. Fig. 1 presents the study area along with species occurrence data retrieved from biodiversity data platforms. The red dots indicate observed occurrence records of Bombus ignitus within the study area. Detailed information on the occurrence data, including sources and collection methods, is provided in Section 2.3.1. Seoul has a mix of artificial and natural environmental factors, offering an ideal setting to analyze the complex interactions between urbanization and species distribution. The species that is the focus of this research, Bombus ignitus, belongs to the order of Hymenoptera and the Apidae family. Bombus ignitus is a key pollinator species in Korea, known for its high mobility and adaptability in both urban and rural environments. Within urban areas, this species is frequently observed in parks, on street trees, and in riparian green spaces, where it forages on a variety of nectar-producing plants. It plays a crucial role in pollination, particularly through interactions with specific plant species that rely on pollinators for reproduction (Park et al., 2018). To comprehensively assess the impact of urbanization on Bombus ignitus, this study considers the entire Seoul metropolitan area as the study boundary. Bombus ignitus has been documented in both central urban zones and peripheral areas, making it essential to capture a full gradient of urbanization effects. However, this species is highly vulnerable to habitat changes and resource scarcity caused by urbanization, which highlights its relevance as a subject for studying the interactions between urban environmental variables and species distribution.

MaxEnt model

The MaxEnt model predicts the potential distribution of species based on occurrence data and environmental variables at those locations. By applying the principle of maximum entropy, it learns the environmental characteristics of areas where a species is present and identifies potential habitats with similar conditions. MaxEnt is known to perform reliably with relatively small datasets, requiring a minimum of 10–20 occurrence records to produce credible results (Pearson, 2007). However, the reliability of the model improves with more occurrence data, reducing the risk of overfitting and increasing the accuracy of spatial predictions (Hernandez et al., 2006). Model performance is typically evaluated using the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) curve. The y-axis of the ROC curve represents the true positive rate, indicating the accuracy of predicting locations where the species is present, while the x-axis shows the false positive rate, representing the proportion of non-habitat areas misclassified as suitable. An AUC above 0.8 is generally considered indicative of strong explanatory power (Elith et al., 2006). Additionally, MaxEnt quantitatively assesses the relative importance and contribution of environmental variables for identifying key factors affecting species distribution. Its ability to perform well with limited data and analyze complex environmental factors, such as urbanization, has led to the wide use of MaxEnt in ecological research (Phillips et al., 2006; Elith et al., 2011).

Input data preparation

Species occurrence data

For this study, 21 occurrence records of Bombus ignitus from 2015 to 2024 were collected from the Global Biodiversity Information Facility (GBIF) website (Fig. 1). GBIF is a global platform that provides free access to biodiversity data, including observational records. Integrating data contributions from researchers, institutions, and citizen scientists worldwide, GBIF serves as a critical resource for research and conservation management. The 21 occurrence records used in this study meet the minimum threshold suggested by Pearson (2007), which supports the use of species distribution modeling with small sample sizes. However, due to the limited number of records, potential constraints associated with data availability were considered when interpreting the results.

Input parameters selection

To analyze the relationship between the presence of Bombus ignitus and urban environmental variables, we conducted a literature review on bumblebee species, including studies that utilized species distribution models for habitat evaluation (Table 1). The variables used in our study were selected based on this review. We first compiled a list of variables commonly used in previous studies, filtered them based on data accessibility from biodiversity and environmental datasets, and assessed their relevance to Bombus ignitus distribution in an urban environment.

Osborne et al. (1999) identified factors influencing bumblebee foraging, such as flight path linearity, direction, forage preference, and range, while also highlighting climatic factors such as wind and temperature. Urban environmental variables, including obstacles to flight paths and land use types, were also considered. Jha and Kremen (2013) emphasized the importance of floral resource diversity and distribution, patch distances, and local land use patterns on foraging and habitat utilization, specifically noting floral diversity as a primary determinant of habitat quality. McFrederick and LeBuhn (2006) identified habitat heterogeneity, floral abundance, and species richness as key factors influencing bumblebee species richness, with connectivity to surrounding green spaces being critical, though park size and age had minimal impact on bumblebee activity. Ahrne et al. (2009) found that floral diversity, plant types, and land boundary lengths significantly influenced bumblebee presence and diversity. Notably, they also observed that an increase in impervious surface areas led to a decline in bumblebee diversity, highlighting the critical role of urbanization in shaping pollinator habitats. These insights guided the selection of variables for this study, ensuring that both natural and urban environmental factors affecting Bombus ignitus were comprehensively analyzed.

Input data preprocessing

The input variables were categorized into three groups: topography and climate, foraging environment, and urban environment. A total of 10 variables were selected (Table 2).

An analysis of the correlations among the 10 selected variables selected based on a review of the literature revealed that the correlation coefficient between plant coverage and vegetation diversity exceeded 0.8, indicating the presence of multicollinearity as shown in Fig. 2. To minimize inaccuracies in the model and clarify interactions among variables, plant coverage was excluded due to its high correlation with vegetation diversity, while the latter was retained for its provision of more detailed information.

For topography and climate, the variables included aspect, temperature, and humidity. For the foraging environment, the variables consisted of plant coverage, actual vegetation, distance from water, and distance from forest. Lastly, for the urban environment, the variables included building coverage, percentage of impervious area, and land cover. These variables were selected to enable a comprehensive analysis of the factors influencing the habitat suitability of Bombus ignitus in urbanized areas.

Climatic data were obtained from the Automatic Weather Station (AWS) of the Korea Meteorological Administration, which provides high-resolution daily records of temperature and humidity. These data were selected to ensure temporal continuity and accuracy for the study. The topographical variable, aspect, was derived from the digital aspect map provided by the Ministry of Environment. This variable was chosen to assess its influence on ecological factors such as solar radiation exposure, temperature variations, and species activity patterns (Jeroen et al., 2011; Choi et al., 2015).

Foraging environment variables, including plant coverage, vegetation diversity, distance from water, and distance from forest, were obtained from the 2020 Seoul biotope map. Plant coverage and vegetation diversity were selected to assess how urban food resources influence the distribution of Bombus ignitus (Weronika et al., 2016). The distances from water and forest were included to analyze the effects of proximity to natural environments on habitat suitability (Weronika et al., 2016).

Urban environment variables included building coverage ratio, percentage of impervious surface area, and land cover. Building coverage ratio and impervious surface area were selected to evaluate the impact of urban development on Bombus ignitus habitats. The land cover variable was incorporated to assess the influence of human activities on species occurrence (Morrison et al., 2012).

For categorical data, the land cover variable was based on the 2023 Land Cover Map published by the Ministry of Environment, which includes 20 subcategories for the study area. The vegetation diversity variable in the model utilized data from the 2020 Seoul biotope map, comprising 9 subcategories. It should be noted that the occurrence data for Bombus ignitus were collected between 2015 and 2024, whereas some environmental variables, such as land cover and vegetation diversity, were derived from datasets from specific years. This temporal mismatch may introduce potential bias, as land cover and vegetation characteristics could have changed over time, particularly in urban environments where large-scale development and green space fragmentation may have occurred.

Multicollinearity among the 10 selected variables was checked using Pearson correlation analysis in R Studio 4.4.1, with the strength of relationships visualized in a correlation matrix. For the species distribution model, 75% of the occurrence data was used for training, and 25% was reserved for validation. Model performance was assessed using the AUC (West et al., 2016). Input data for the MaxEnt model were prepared in raster format with a spatial resolution of 30 m × 30 m, using QGIS 3.35.1 and ArcGIS Pro 3.0.2, as shown in Fig. 3.

Habitat suitability analysis

To analyze how urban environmental variables influence the potential habitat distribution of Bombus ignitus, two scenarios were developed using MaxEnt version 3.4.4 (Phillips et al., 2006). Scenario 1 included urban factor in the model’s environmental variables, while Scenario 2 excluded urban variables, focusing only on the remaining environmental factors (topography, climate, and foraging). This approach allowed for a comparative analysis of the impact of urban environmental variables on habitat suitability. The results obtained using both models were compared and visualized spatially to quantitatively assess the impact of urban environmental variables on habitat suitability. The entire process is illustrated in Fig. 4.

To quantitatively analyze the differences in habitat suitability between the two models, habitat suitability values, which range from 0 to 1, were categorized into four levels based on the likelihood of species occurrence. Higher values indicate a greater likelihood of presence, with values between 0.75 and 1 classified as ‘Optimal,’ between 0.5 and 0.74 as ‘Suitable,’ between 0.25 and 0.49 as ‘Less suitable’ and between 0 and 0.24 as ‘Unsuitable.’ This classification allowed for a structured comparison of habitat suitability patterns under scenarios with and without the inclusion of urban environmental variables.

Results and Discussion

Comparison of potential habitat distribution results

The potential habitat distribution of Bombus ignitus was predicted using two scenarios: Scenario 1 included urban environmental variables, while Scenario 2 excluded them. Fig. 5 shows the ROC curves used to evaluate the predictive performance of each model. The AUC value for Scenario 1 was 0.916, while for Scenario 2 it was 0.913. Both models demonstrated high predictive performance, with AUC values exceeding 0.9, regardless of whether urban environmental variables were included. However, the slightly higher AUC value for Scenario 1 indicates a marginal improvement in predictive performance when urban environmental variables were included. These results suggest that urban environmental variables contribute to explaining the distribution of Bombus ignitus. Nevertheless, the minimal difference in AUC values implies that urban variables play a relatively minor role in species distribution prediction. Overall, both models exhibited strong predictive reliability, with the inclusion of urban environmental variables having a limited impact on model performance. This finding suggests that key factors influencing the distribution of Bombus ignitus are likely natural environmental elements, such as climate and food resources, rather than urbanization-related variables. Further research integrating field data and additional environmental factors is needed to gain a more comprehensive understanding of these interactions. Notably, additional field surveys and independent validation efforts are necessary to enhance the reliability of the model’s predictions and ensure a more robust assessment of Bombus ignitus habitat suitability.

The results of the habitat suitability analysis for the two scenarios are illustrated in Fig. 6, and show minimal differences between the two scenarios. As presented in Table 3, the area proportions of the four habitat suitability categories across Seoul illustrate that ‘Unsuitable’ areas comprised 72.93% in Scenario 1 and 76.96% in Scenario 2. The proportions for the ‘Optimal,’ ‘Suitable’ and ‘Less suitable’ categories were 4.83%, 5.38%, and 16.87% in Scenario 1, respectively, compared to 4.29%, 4.94%, and 13.82% in Scenario 2, respectively. These results indicate that urban environmental variables have a limited effect on the distribution of habitat suitability. One possible explanation for the minimal differences between the two scenarios is that the urban environmental variables added in Scenario 2 might have relatively low explanatory power compared to the natural environmental factors already considered in Scenario 1. Additionally, some of the urban variables may be spatially correlated with natural variables, resulting in overlapping explanatory effects. As a result, the overall habitat suitability patterns remained largely unchanged despite the inclusion of urban environmental variables. This suggests that natural environmental factors play a more dominant role in determining habitat suitability in Seoul, while urban environmental conditions may contribute only marginally.

Fig. 7 illustrates areas categorized as ‘Optimal’ for habitat suitability (suitability range: 0.75–1). These locations include regions close to forests, small green spaces near the Han River, and major urban green zones such as ‘Northern Seoul Dream Forest,’ ‘Blue Arboretum’ and ‘West Seoul Lake Park.’ In Scenario 2, even without urban environmental variables, regions near forests and water bodies remained key suitable habitats, reaffirming the importance of natural factors. It also highlights the fact that forested areas and large mountainous regions within Seoul at elevations below 100 meters exhibit high habitat suitability for Bombus ignitus. This includes major forested regions such as Mt. Bukak, Mt. Gwanak, and Mt. Nam, as well as smaller urban green spaces, indicating their role in supporting pollinator habitats. Notably, Nodeul Island along the Han River shows high suitability, suggesting that even within urban settings, specific ecological conditions create favorable habitats.

Additionally, the ecological characteristics of the six locations with optimal habitat suitability highlighted in Fig. 7 are summarized as follows. Location 1 (Mt. Sumyeong) is a green corridor along the Han River with natural patches that enhance habitat connectivity and provide diverse floral resources for foraging and nesting. Location 2 (Bamseom Island and Nodeul Island) is two river islands with semi-natural habitats, minimal human disturbance, and abundant flowering plants, offering highly suitable conditions for pollinators. Location 3 (Mt. Namsan), a centrally located urban forest, supports pollinators with its large green space and rich floral diversity. Notably, Nodeul Island benefits from diverse herbaceous vegetation, limited human disturbance, and its location within the Han River corridor, which enhances connectivity to nearby green spaces and supports stable pollinator habitats. Location 4 (Mt. Gwanaksan) is a vast forested area with continuous natural vegetation and low elevation, providing stable and diverse habitats. Location 5 (Jungnang Stream) is a linear riparian corridor with adjacent green spaces and aquatic vegetation, functioning as an important ecological pathway for pollinators. Location 6 (Mt. Cheonggyesan) is a south-eastern mountainous forest with rich native flora and relatively low urban disturbance, serving as a key habitat refuge for Bombus ignitus. These areas collectively highlight the importance of preserving forests, river islands, and riparian corridors to maintain pollinator habitats in urban environments.

Response curve analysis

The response curves for the variables in Scenario 1, which includes urban environmental factors, are shown in Fig. 8. The y-axis of each response curve represents the predicted probability of species presence for a given environmental variable. This reflects a relative pattern rather than absolute probability, as all other environmental variables are held constant at their average values. This approach simplifies the interpretation of each variable's effect.

For climatic and topographical variables, the aspect (Fig. 8(a)) indicated the highest probability of presence in northwestern areas. This can be attributed to the reduced direct solar radiation in northwestern areas compared to southern areas, which helps mitigate excessive temperature increases during summer. During spring and summer, when Bombus ignitus is most active, cooler areas may provide more favorable conditions. The temperature (Fig. 8(b)) showed the highest probability of presence at 22°C, with a decline in probability as temperatures increased. This result aligns with the known thermal response of Bombus ignitus, which shows stable activity within the 10–20°C range, although peak activity may occur at slightly higher temperatures, around 22–23°C.

In terms of humidity (Fig. 8(c)), there were no significant changes in the probability of presence. This suggests that while humidity might not be a primary driver of Bombus ignitus distribution, it could still have indirect effects through interactions with other key variables in the model. For the distance from water (Fig. 8(d)), the probability of presence increased as the distance from water bodies grew. This suggests that Bombus ignitus primarily relies on plants for food resources, which reduces the importance of proximity to water. Additionally, the Han River, the major water body in Seoul, traverses the city from west to east. The avoidance of artificial environments, such as buildings near the water, might further explain this pattern.

In contrast, the probability of presence sharply declined as distance from forests increased (Fig. 8(e)). As Bombus ignitus typically nests in underground burrows, tree holes, or decayed wood, forests provide essential natural resources critical for its survival and activity (Werner et al., 2014). Habitat fragmentation due to urbanization and forest isolation might negatively affect its distribution, highlighting the importance of forest conservation in urban and natural environment management. For vegetation diversity (Fig. 8(f)), the probability of presence was highest in areas such as landscaped tree plantations (category 3), bodies of water (category 5), and forested areas (category 9). Among these, landscaped tree plantations exhibited a higher probability than grasslands or natural forests, likely due to their diverse vegetation and multi-layered structure, which may make them a preferred habitat for Bombus ignitus.

For land cover (Fig. 8(g)), the highest probabilities of presence were observed in farmland (category 8), natural bare land (category 19), coniferous forests (category 13), artificial grasslands (category 16), and broad-leaved forests (category 12). The preference for farmland aligns with previous findings indicating that Bombus ignitus is more efficient than honeybees in pollinating crops (Velthuis and Doorn, 2006). Farmlands provide a rich and continuous supply of floral resources, particularly crops that require pollination, contributing to its high suitability in agricultural areas. The relatively high suitability observed in natural bare land (category 19) can be attributed to the presence of open habitats with early successional vegetation, which offers diverse floral resources and nesting opportunities in well-drained, sun-exposed soils preferred by Bombus ignitus. Similarly, coniferous forests (category 13) provide relatively undisturbed, stable habitats with diverse understory vegetation, creating suitable nesting and foraging conditions. Artificial grasslands (category 16), which include urban parks and green spaces with maintained vegetation, also exhibited moderate habitat suitability. These areas offer fragmented but valuable floral resources that can support pollinators, particularly in the highly urbanized landscapes of Seoul.

In conclusion, the land cover variable significantly influences the habitat selection of Bombus ignitus, with agricultural areas and farmland playing a particularly crucial role. This underscores the species’ ecological importance not only in supporting natural pollination processes but also in enhancing crop productivity in agricultural landscapes. These findings highlight the need for habitat conservation strategies that account for the ecological contributions of Bombus ignitus across both natural and human-modified environments, particularly in agroecosystems, where its pollination services are essential.

The analysis of urban environmental variables, such as building coverage ratio and impervious surface area (see Fig. 8(h) and Fig. 8(i)), revealed a consistent decrease in the probability of the presence of Bombus ignitus as building coverage approached 100%. Similarly, the probability of presence declined sharply when impervious surface area coverage exceeded approximately 70%. These patterns indicate that Bombus ignitus tends to avoid habitats significantly altered by human activities. Changes associated with urbanization, including reduced vegetation, limited food resources, and the lack of suitable nesting sites, are key factors that restrict habitat suitability for Bombus ignitus in areas with high impervious surface ratios. The findings underscore the substantial negative impact of building coverage and impervious surfaces on habitat suitability and highlight the importance of these variables in urbanization studies.

In conclusion, building coverage and impervious surface area have a detrimental effect on the habitat suitability of Bombus ignitus. Urban planning and management strategies should focus on preserving green spaces and open areas while minimizing impervious surfaces to conserve Bombus ignitus habitats and promote biodiversity in cities.

Variable contribution and importance results

Table 4 summarizes the contribution percentages and permutation importance of key variables that affect the predicted habitat suitability of Bombus ignitus. Contribution indicates the extent to which each variable explains habitat suitability during the model training process. Initially, contributions are either evenly distributed or randomly assigned, but they are gradually adjusted as the model assesses the impact of each variable while minimizing the influence of correlated ones. Permutation importance measures the independent significance of each variable after the model training is complete. This is done by randomizing the values of a variable and observing the resulting changes in the AUC. A larger drop in AUC indicates a higher permutation importance for that variable.

In scenario 1, which includes urban environmental variables, the three most significant contributing factors were temperature (37.02%), distance from forests (27.75%), and land cover (13.34%). According to permutation importance, the leading variables were distance from forests (44.14%), temperature (21.66%), and vegetation diversity (15.46%). In Scenario 2, which excludes urban environmental variables, the most contributing factors were temperature (42.28%), distance from forests (29.46%), and distance from water (12.88%). The permutation importance ranking indicated that temperature (45.78%), distance from forests (37.77%), and distance from water (13.02%) were the top variables.

Both scenarios emphasize that temperature and distance from forests are critical determinants of Bombus ignitus habitat suitability. Temperature is a key climatic factor influencing the foraging behavior, nesting success, and overall colony development of Bombus ignitus. This reflects the species’ physiological sensitivity to thermal conditions, which influence metabolic rates, foraging efficiency, and larval development (Rasmont et al., 2015).

Distance from forests emerged as a consistently important factor in both scenarios. Forest edges provide critical nesting habitats, diverse floral resources, and shelter from extreme environmental conditions, which are essential for supporting Bombus ignitus populations. Additionally, forest edges enhance habitat connectivity, allowing pollinators to move between resource patches more effectively (Goulson et al., 2008). These findings align with those of previous studies, highlighting that maintaining natural habitat connectivity is vital for supporting bumblebee populations.

Urban environmental variables demonstrated relatively low contributions, indicating they have a limited influence on habitat suitability in this study area. Similar patterns were observed in previous studies, where natural factors such as vegetation cover and proximity to green spaces played a far more critical role in determining pollinator distributions in urban landscapes (Hall et al., 2017). However, previous studies have also suggested that well-managed urban green spaces, such as parks and gardens, can provide complementary resources for pollinators within highly urbanized landscapes, indicating that urban variables may play a more localized role under specific conditions.

Specifically, the ‘aspect’ variable showed low contribution and permutation importance in both scenarios, suggesting that Bombus ignitus is not particularly sensitive to specific orientations, or that aspect has a minor role in habitat selection. These findings imply that natural factors, such as climate and proximity to forests, should be prioritized over anthropogenic factors such as urbanization when planning and implementing Bombus ignitus conservation strategies. At the same time, enhancing urban green spaces and improving habitat connectivity within cities could provide additional support for pollinator populations, especially in rapidly urbanizing regions. This information could provide essential baseline data for developing effective biodiversity conservation strategies that integrate both natural and urban ecosystems.

Conclusion

This study utilized the MaxEnt model to compare and analyze potential changes in the distribution of Bombus ignitus, both with and without urban environmental variables. It also examined how these selected environmental variables impact habitat suitability. Both the model that included urban variables (Scenario 1) and the model that excluded them (Scenario 2) demonstrated high predictive performance, with minimal differences in habitat suitability distribution observed between the two scenarios. This suggests that urban environmental variables are not the primary determinants of habitat suitability for Bombus ignitus. However, the analysis of variable contributions and response curves indicated that interactions between urban and natural environmental variables could influence habitat suitability. Among the urban variables, building coverage and impervious surface area were found to decrease the probability of Bombus ignitus presence, suggesting that urbanization could limit habitat availability under unfavorable conditions. Conversely, natural variables such as distance from forests and temperature were identified as critical determinants of habitat suitability. Forests, in particular, provide essential food resources and nesting spaces, playing a significant role in maintaining habitat connectivity even within urban environments. Additionally, landscaped green spaces such as tree plantations were evaluated as favorable habitats for Bombus ignitus, demonstrating their potential to enhance urban biodiversity and pollination services. These findings underscore the importance of incorporating both natural and urban environmental considerations in conservation and management strategies to conserve Bombus ignitus populations and promote ecosystem services in urban settings. In particular, it is necessary to improve landscape policies to facilitate the planning of sustainable urban ecosystems. These policies should focus on increasing the implementation of Nature-based Solutions (NbS) and promoting activities to enhance ecosystem services, such as expanding and connecting urban green spaces.

The main implications of this study are as follows. First, high-quality green spaces such as landscaped tree planting sites play a critical role in the habitat of Bombus ignitus, emphasizing the need for strategies to expand and enhance the connectivity of such spaces in urban planning processes. Second, urban environmental variables such as building coverage ratios and impervious surface ratios negatively affect the habitat suitability of Bombus ignitus. However, they were not identified as major determinants, indicating the need for a complementary approach that considers interactions between urbanization and natural environmental factors. Third, beyond simple correlations between environmental variables and species distribution, an integrated modeling approach is required to reflect the complex interactions between the ecological needs of Bombus ignitus and its environment.

While this study provides valuable insights through modeling, it has several limitations. Linking habitat suitability to environmental variables across a large urban area such as Seoul inherently simplifies the spatial heterogeneity of microhabitats and localized ecological processes. In particular, while the environmental variables used in the model effectively capture broad-scale trends, they may not fully represent localized microclimatic conditions or species-specific resource preferences.

Additionally, the occurrence data from GBIF, although widely used and reliable, could introduce spatial biases due to data gaps, uneven sampling effort, or duplicate records. The relatively small number of species occurrence records within Seoul may also have affected the robustness of the model, limiting its ability to fully capture the diverse environmental conditions associated with Bombus ignitus habitats. Therefore, we interpreted the results with caution, taking into account potential biases in data collection and geographic coverage.

To address these limitations, future studies should incorporate field surveys to gain an understanding of Bombus ignitus’s habitat use and foraging behavior. Incorporating more extensive field survey results based on the modeling approach presented in this study could further improve the model predictions for insect distribution in future research. This integrated approach would also help identify key ecological factors driving habitat suitability, informing more targeted and effective pollinator conservation strategies in urban ecosystems.

Notes

This study was supported by a 2024 Research Grant (Project number: 202403800001) from Kangwon National University.

This work was supported by a National Institute of Biological Resources (NIBR) grant funded by the Ministry of Environment (MOE) of the Republic of Korea (Project number: 202501490001).

Fig. 1

Spatial distribution of occurrence points for Bombus ignitus within the study area (Seoul). The red dots represent observation points collected from the GBIF database (2015–2024).

Fig. 2

Correlation coefficients between variables, visualized using a correlation matrix. Blue represents positive correlations, while red indicates negative correlations. The size of the circles corresponds to the absolute value of the correlation coefficient, with larger circles indicating stronger correlations.

Fig. 3

Environmental variables used in the MaxEnt model included aspect, temperature, humidity, distance from water, distance from forest, actual vegetation, land cover, building coverage, and percentage of impervious area.

Fig. 4

A flowchart for predicting Bombus ignitus habitat suitability using MaxEnt model.

Fig. 5

ROC curve showing the AUC for the prediction species distribution incorporating urban environmental variables.

Fig. 6

Habitat suitability map for Scenario 1, which included urban environmental variables (A), and Scenario 2, which excluded urban environmental variables.

Fig. 7

Spatial distribution of optimal habitat suitability for Bombus ignitus in scenario 1. This map visualizes the predicted optimal habitat areas (suitability range: 0.75–1) for Bombus ignitus within Seoul, highlighting key regions such as forests, green spaces, and parks. Locations numbered 1 to 6 represent specific high-suitability zones (1: Mt. Sumyeong, 2: Bamseom Island and Nodeul Island, 3: Mt. Namsan, 4: Mt. Gwanaksan, 5: Jungnang Stream, 6: Mt. Cheonggyesan).

Fig. 8

Response curves for scenario 1 including urban environmental variables. The y-axis of each response curve represents the predicted probability of species presence for a given environmental variable: (a) Aspect, (b) Temperature, (c) Humidity, (d) Distance from water, (e) Distance from forest, (f) Actual vegetation, (g) Land cover, (h) Building coverage, (i) Percentage of impervious area.

Table 1

Summary of variables related to foraging, topography, climate, and urban environment from previous studies on bee ecology

References	Categories (Variable)	Criteria
Osborne et al. (1999)	Foraging	Flight speed and straightness of foraging routes
		Flight direction
		Forage preference
		Foraging range
		Pollen load presence

	Topography & Climate	Temperature
	Topography & Climate	Wind speed

	Urban Environment	Flight path obstructions
	Urban Environment	Land use

Jha and Kremen (2013)	Foraging	Floral cover
		Foraging distance
		Species richness

	Urban Environment	Landscape (among patches) scale
		Local (within patch) scale
		Land cover

McFrederick and LeBuhn (2006)	Foraging	Habitat heterogeneity
		Abundance of a dominant bumblebee species
		Number of rodent holes
		Floral abundance
		Floral species richness

	Topography & Climate	Mean elevation

	Urban Environment	Edge to area ratio
		Park area and age
		Proportion of open space in the surrounding area (Openness of the surrounding matrix)
		Distance of each park from the nearest possible source population

Ahrne et al. (2019)	Foraging	Flower cover
		Flower richness
		Number of flowering plant species

	Urban Environment	Allotment age
		Boundary lengths by land cover type
		Land cover
		Type of site (house, garden, mixed)

Table 2

Description of topographic, climatic, foraging, and urban environment variables used for the MaxEnt model

Categories	Variables (Abbreviation)	Unit	Raw data (year)	Data Source (year)
Topography & Climate	Aspect (A)	Degree	Digital topographic map (2022)	Ministry of Land, Infrastructure and Transport
	Temperature (T)	°C	Daily mean temperature data (2015–2024)	Korea Meteorological Administration
	Humidity (H)	%	Daily mean temperature data (2015–2024)	Korea Meteorological Administration
Foraging	Plant coverage (PC)	%	Seoul biotope map (2020)	Seoul Development Institute
	Actual vegetation (AV)	-
	Distance from water (DW)	m
	Distance from forest (DF)	m
Urban Environment	Landcover (LC)	-	Land cover map (2023)	Ministry of Environment
	Building coverage (BC)	%	Seoul biotope map (2020)	Seoul Development Institute
	Percentage of impervious area (I)	%	Seoul biotope map (2020)	Seoul Development Institute

Table 3

Habitat suitability legends and their area proportions

Legend (Habitat suitability)	Optimal	Suitable	Less suitable	Unsuitable
Scenario 1 (A)	4.29%	4.94%	13.82%	76.96%
Scenario 2 (B)	4.83%	5.38%	16.87%	72.93%
B-A	0.54%	0.44%	3.05%	−4.02%

Table 4

Contribution and permutation importance of environmental variables in scenario 1 and scenario 2. The variables are arranged in descending order based on their contribution and permutation importance

	Variables	Contribution (%)	Permutation importance (%)
Scenario 1	Temperature	37.02	21.66
	Distance from forest	24.75	44.14
	Land cover	13.34	12.79
	Actual vegetation	11.93	15.46
	Distance from water	8.98	5.59
	Aspect	3.83	0
	Percentage of impervious area	0.12	0
	Building coverage	0.03	0.36
	Humidity	0	0

Scenario 2	Temperature	42.28	45.78
	Distance from forest	29.46	37.77
	Distance from water	12.88	13.02
	Actual vegetation	11.12	3.43
	Aspect	4.26	0
	Humidity	0	0

References

Ahrne, K., J. Bengtsson, T. Elmqvist. 2009. Bumble bees (Bombus. spp) along a gradient of increasing urbanization. PloS one. 4(5):e5574.https://doi.org/10.1371/journal.pone.0005574

Barth, F.G. 1985. Insects and flowers. The Biology of a Partnership George Allen and Unwin. United Kingdom:

Choi, Y.S., M.Y. Lee, I.P. Hong, N.S. Kim, H.K. Kim. 2010. Detection of honeybee virus from bumblebee (Bombus terrestris L. and Bombus ignitus). Journal of Apiculture. 25(4):259-266.

Choi, Y.S., M.Y. Lee, M.L. Lee, H.K. Kim, N.S. Kim, I. PHong, K.H. Byeon, K.G. Lee, H.J. Yoon. 2009. Detection of deformed wing virus (DWV) from bumblebee (Bombus ignitus) in Korea. Korean Journal of Apiculture. 24(2):83-91.

Elith, J., H.C. Graham, P.R. Anderson, M. Dudík, S. Ferrier, A. Guisan, E.N. Zimmermann. 2006. Novel methods improve prediction of species’ distributions from occurrence data. Ecography. 29(2):129-151. https://doi.org/10.1111/j.2006.0906-7590.04596.x

Elith, J., S.J. Phillips, T. Hastie, M. Dudík, Y.E. Chee, C.J. Yates. 2011. A statistical explanation of MaxEnt for ecologists. Diversity and Distributions. 17(1):43-57.

Franklin, J. 2010. Mapping species distributions: spatial inference and prediction Cambridge University Press.

Gill, N.S., F. Sangermano. 2016. Africanized honeybee habitatsuitability: a comparison between models for southern Utah and southern California. Applied Geography. 76:14-21.

Goulson, D. 2003. Bumblebees: their behaviour and ecology Oxford University Press. England:

Goulson, D., G.C. Lye, B. Darvill. 2008. Decline and conservation of bumble bees. Annual Review of Entomology. 53(1):191-208. https://doi.org/10.1146/annurev.ento.53.103106.093454

Hall, D.M., G.R. Camilo, R.K. Tonietto, J. Ollerton, K. Ahrné, M. Arduser, C.G. Threlfall. 2017. The city as a refuge for insect pollinators. Conservation Biology. 31(1):24-29.

Hernandez, P.A., C.H. Graham, L.L. Master, D.L. Albert. 2006. The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography. 29(5):773-785.

Hevia, T., J. Bosch, F.M. Azcárate, E. Fernández, A. Rodrigo, H. Barril-Graells, J.A. González. 2016. Bee diversity and abundance in a livestock drove road and its impact on pollination and seed set in adjacent sunflower fields. Agriculture, Ecosystems and Environment. 232:336-344.

Jeroen, E., W. Michael, G. Bernd, F. Carsten. 2011. Microsite conditions dominate habitat selection of the red mason bee (Osmia bicornis, Hymenoptera: Megachilidae) in an urban environment: A case study from Leipzig, Germany. Landscape and Urban Planning. 103:15-23.

Jha, P.K., R.B. Thapa, J.B. Shrestha. 2005. Conservation and Management of Pollinators for Sustainable Agriculture through an Eco-system Approach Report submitted to Food and agriculture Organization of the United Nations. Kathmandu:

Jha, S., C. Kremen. 2013. Resource diversity and landscape-level homogeneity drive native bee foraging. Proceedings of the National Academy of Sciences. 110(2):555-558.

Kim, Y.H. 2021. Predicting the appearance of Apis mellifera and Bombus ignitus for the promotion of urban ecosystem pollination servic. Master's thesis. Kongju National University Graduate School, Gongju, Korea.

Lee, S.B., T.W. Bae, H.J. Yoon. 2005. The effect of light on the colony development of Korean native bumblebee, Bombus ignitus Smith (Hymenoptera: Apidae). Journal of Apiculture. 20(1):67-76.

Lee, S.H., L. Dumouchel. 1999. Taxanomic review of Genus Bombus (Hymenoptera, Apidae) from Korea. Ins. Koreana. 16:77-101.

Martins, K.T., A. Gonzalez, M.J. Lechowicz. 2015. Pollination services are mediated by bee functional diversity and landscape context. Agriculture, Ecosystems and Environment. 200:12-20. https://doi.org/10.1016/j.agee.2014.10.018

Matteson, K.C., J.S. Ascher, G.A. Langellotto. 2008. Bee richness and abundance in New York City urban gardens. Annals of the Entomological Society of America. 101(1):140-150.

Matteson, K.C., G.A. Langellotto. 2010 Determinates of inner city butterfly and bee species richness. Urban ecosystems 13:333-347. retrieved from https://link.springer.com/article/10.1007/s11252-010-0122-y.

McFrederick, Q.S., G. LeBuhn. 2006. Are urban parks refuges for bumble bees Bombus spp.(Hymenoptera: Apidae)? Biological conservation. 129(3):372-382.

McKinney, M.L. 2008. Effects of urbanization on species richness: a review of plants and animals. Urban ecosystems. 11(2):161-176.

Min, S.H., J.M. Kim, S.J. Lim, B.H. Kim, C.W. Lee, B.S. Yoon. 2017. Development of real-time polymerase chain reaction detection method for 11 pathogens of bumblebees. Korean Journal of Apiculture. 32(2):99-109. https://doi.org/10.17519/apiculture.2017.06.32.2.99

Morandin, L.A., M.L. Winston. 2005. Wild bee abundance and seed production in conventional, organic, and genetically modified canola. Ecological Applications. 15(3):871-881.

Morrison, M.L., B. Marcot, W. Mannan. 2012. Wildlife-habitat relationships: concepts and applications Island Press. Washington · Covelo · London:

Nabhan, G.P., S.L. Buchmann. 1997. Services provided by pollinators. Nature's Services: societal dependence on natural ecosystems. 133-150.

Osborne, J.L., S.J. Clark, R.J. Morris, I.H. Williams, J.R. Riley, A.D. Smith, D.R. Reynolds, A.S. Edwards. 1999. A landscape-scale study of bumble bee foraging range and constancy, using harmonic radar. Journal of Applied Ecology. 36(4):519-533.

Park, I.G., H.J. Yoon, S.B. Lee, S.E. Kim, Y.I. Lee. 2004. Characteristics of oviposition and colony development and preference of the bumblebee, Bombus ignitus, with several concentrations of sugar solution. Journal of Apiculture. 19(2):83-88.

Park, J., D.S. Kim, K.H. Song, T.J. Jeong, S.J. Park. 2018. Mapping potential habitats for the management of exportable insects in South Korea. Journal of Asia-Pacific Biodiversity. 11(1):11-20.

Pearson, R.G. 2007. Species’ distribution modeling for conservation educators and practitioners. Synthesis/American Museum of Natural History. 50.

Phillips, S.J., R.P. Anderson, R.E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling. 190(3–4):231-259.

Potts, S.G., J.C. Biesmeijer, C. Kremen, P. Neumann, O. Schweiger, W.E. Kunin. 2010. Global pollinator declines: trends, impacts and drivers. Trends in ecology and Evolution. 25(6):345-353. https://doi.org/10.1016/j.tree.2010.01.007

Rasmont, P., M. Franzen, T. Lecocq, A. Harpke, S.P. Roberts, J.C. Biesmeijer, O. Schweiger. 2015. Climatic risk and distribution atlas of European bumblebees. Bio Risk. 10:1-236.

Song, W.K., E.Y. Kim. 2012. A comparison of machine learning species distribution methods for habitat analysis of the Korea water deer (Hydropotes inermis argyropus). Korean Journal of Remote Sensing. 28(1):171-180. https://doi.org/10.7780/kjrs.2012.28.1.171

Vallejo-Marín, M. 2019. Buzz pollination: studying bee vibrations on flowers. New Phytologist. 224(3):1068-1074.

Velthuis, H.H., A. Van Doorn. 2006. A century of advances in bumblebee domestication and the economic and environmental aspects of its commercialization for pollination. Apidologie. 37(4):421-451. https://doi.org/10.1051/apido:2006019

Weronika, B., R. Halina, D. Lukasz. 2016. Features of urban green space favourable for large and diverse bee populations (Hymenoptera: Apoidea: Apiformes). Urban Forestry and Urban Greening. 20:448-452. https://doi.org/10.1016/j.ufug.2016.10.015

West, A.M., S. Kumar, C.S. Brown, T.J. Stohlgren, J. Bromberg. 2016. Field validation of an invasive species Maxent model. Ecological informatics. 36:126-134. https://doi.org/10.1016/j.ecoinf.2016.11.001

Yoon, H.J., S.E. Kim. 2002. Facilitating effects of helpers on oviposition and colony development of bumblebee queen, Bombus ignitus. Korean Journal of Applied Entomology. 41(4):239-245.

Yoon, H.J., H.S. Shim, Y.S. Kim, S.E. Kim, S.B. Lee. 2005. Light source, temperature, and age favorable for mating of the bumblebee, Bombus ignitus and B. terrestris. Journal of Apiculture. 20(1):59-66.

Yoon, H.J., K.Y. Lee, H.J. Ko. 2018. Sexual maturity time of reproductive organ development and mating in the Korean native bumblebee, Bombus ignitus. Korean Journal of Applied Entomology. 57(4):329-337. https://doi.org/10.5656/KSAE.2018.10.0.050

Yoon, H.J., S.E. Kim, S.B. Lee, I.G. Park. 2003 Effect of CO₂-treatment on oviposition and colony development of the bumblebee, Bombus ignitus. Korean Journal of Applied Entomology 42(2):139-144. retrieved from https://uci.or.kr/I410-ECN-0102-2009-490-003392583.

Yoon, H.-J., S.-E. Kim, Y.-S. Kim, S.-B. Lee, I.-G. Park. 2005. Comparison of colony development of the Korean native bumblebee, bombus ignitus and B. ardens depending on pasturage time and place at field net house. Journal of Apiculture. 20(2):133-142.