Introduction

The investigation of endemic plant species, defined as flora exclusive to, often secluded areas, is pivotal for global biodiversity conservation efforts1. In particular, the endemic flora of the Korean Peninsula stands out as a unique fragment of global vegetation, unmatched anywhere else2. These plants play a crucial role in sustaining ecological equilibrium and preserving the distinctive features of their habitats3,4​​. Their conservation transcends ecological interests, entering the realm of national bio-sovereignty, particularly highlighted by the Nagoya Protocol under the Convention on Biological Diversity (CBD), which stresses the importance of Access and Benefit-Sharing (ABS) of genetic resources. Endemic species are prized for their unique genetic compositions and adaptations to distinct ecological niches, making them invaluable assets.

Beyond their local significance, endemic plants play a crucial role in global biodiversity. Their specialized nature and adaptation to specific environmental conditions make them highly vulnerable to extinction, posing a significant global concern that diminishes Earth’s biodiversity5. This vulnerability is further intensified by the current environmental crisis, particularly climate change, which has been shown to cause immediate and substantial biodiversity alterations6,7. These changes include geographic range shifts, alterations in phenology, and physiological adjustments, among other effects8,9,10,11,12.

Species’ migration, characterized by the expansion or contraction of their geographic ranges, serves as a primary response to climate change13. This response is influenced by several factors, including geographic range size, ecological adaptability, and interactions with other species and the environment14. Endemic plants, with their limited ranges and specialized adaptations, are especially at risk15,16,17. Their limited dispersal capabilities, small population sizes, and lower adaptive capacities render them particularly susceptible to localized threats, which are likely to increase with climate change5. A meta-analysis involving 232 studies from 273 biodiversity hotspots globally predicts that terrestrial endemic plants are disproportionately affected by climate change compared to non-endemic and introduced species7. Endemics in isolated islands or mountainous areas face even greater risks18. These findings highlight the vulnerability of endemic plants to climate change, underscoring the urgency of their conservation. Given these challenges, it is paramount to prioritize the conservation of endemic plants within national conservation strategies. These strategies should encompass not only habitat protection but also proactive measures like habitat restoration and population relocation. Early monitoring for climate change impacts is essential for timely interventions19. However, crafting effective conservation plans for the diverse array of endemic species, each with its unique ecological needs, is complex and further complicated by constraints such as limited resources.

To overcome these limitations, it is crucial to prioritize conservation efforts based on the vulnerability of each species or functional group to climate change20. This approach necessitates the development of spatial data that can predict changes in the abundance and distribution of these plants under various climate change scenarios. Such data are essential for identifying and conserving core biodiversity areas, thereby exerting a protective effect on surrounding habitats. In this respect, recent vulnerability studies of species’ geographic responses to climate change have shifted from individual analyses of small numbers of species to studies of the geographic responses of multispecies groups comprising hundreds or more species. These studies aim to understand the complex effects of climate change on global biodiversity hotspots by examining how species might adapt or shift their ranges in response to changing environmental conditions​11,21. The research primarily focuses on endemic species, providing crucial insights into their potential distribution and survival, which are essential for developing conservation strategies21,22,23​​. This approach underscores the necessity for proactive measures to mitigate the long-term impacts of climate change on biodiversity11,23​​. In Korea, most studies on the geographic response of species are limited to studies on a few species based on species distribution modelling24,25,26,27, and modelling of changes in biodiversity by multispecies groups is extremely rare28. This gap highlights the need for a multispecies group-based climate change vulnerability assessment that reflects the unique characteristics of the Korean Peninsula.

Due to limited access to data on endemic species in the North Korea, and the lack of validation of their original habitats through platforms like GBIF (Global Biodiversity Information Facility), we focus our analysis primarily on endemic plants native to the South Korea. However, given that the Korean Peninsula is a continuous landmass and climate change is expected to cause a northward shift in species distributions29, restricting our model to the South Korea may be scientifically limiting. Therefore, we modeled the entire peninsula to provide comprehensive and scientifically robust projections.

This study aims to fill this gap by focusing on three main objectives: (1) deriving current habitat characteristics of each South Korean endemic plant species using spatial statistical analysis and species distribution modeling, (2) identifying changes in distribution areas in response to future climate change scenarios, and (3) Applying indicators to assess the vulnerability of special plants to climate change, based on the analyzed results. This approach will enable a comprehensive understanding of distributional changes in these species due to climate shifts, based on predictive models. Consequently, this study seeks to make significant contributions to national biosecurity and global biodiversity conservation, offering new perspectives and methodologies for protecting these invaluable natural resources.

Materials and methods

Study area

The Korean Peninsula, central to our study, is distinguished not only by its remarkable biodiversity and unique ecosystem dynamics but also by its complex vegetation climate zone characteristics (Fig. 1). Spanning approximately 1,100 km from north to south and bordered by the Yellow Sea, the East Sea, and the Korea Strait, the region’s landscape is predominantly shaped by the Baekdu-daegan mountain range. This range, incorporating the Baekdu and Jiri mountains, forms the ecological backbone of the peninsula, serving as a natural climatic barrier and a refuge for numerous endemic plant species, thereby playing a critical role in biodiversity conservation30.

Fig. 1
figure 1

Overview map of the Korean Peninsula showing the Baekdu-daegan mountain range, major mountains, and islands. This study includes both South and North Korea to highlight the potential northward movement of South Korean endemic species. The map was generated using ArcGIS Pro version 3.2 (Esri Inc., https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview). Map data sources include Esri South Korea, TomTom, Garmin, FAO, NOAA, and USGS. The map complies with Esri’s non-commercial use policy for academic research.

Full size image

Characterized by a temperate climate with pronounced seasonal variations, the peninsula experiences cold, dry winters, and hot, humid summers. During the period from 1991 to 2020, average temperatures in the coldest month dropped to as low as – 12.3 °C in the coldest regions, while in the hottest month, temperatures rose to as high as 26.7 °C in the hottest regions31. Precipitation is heavily seasonal, predominantly occurring during the monsoon season from June to September, shaping the region’s diverse habitats and ecosystems31. Recent climate data indicate an average temperature increase of approximately 1.5 °C over the past century, with significant variations between the northern and southern regions32. Projections suggest further temperature increases and altered precipitation patterns, with potentially profound impacts on local ecosystems33.

The peninsula’s geological framework is marked by the ancient and stable Baekdu-daegan mountain range, mainly composed of Precambrian rocks such as gneiss and granite. This geological diversity underpins a variety of soil types across the region, each fostering unique ecosystems. The rugged terrain significantly influences the peninsula’s hydrology and watershed dynamics, further affecting local biodiversity34.

Land use across the Korean Peninsula reflects a dynamic interplay between urbanization, agriculture, and ecological conservation. Urban areas are concentrated in lower elevations and coastal plains, while agricultural activities dominate the central and southern regions, focusing on crops like rice and barley. The lowland forests are dominated by plantations of Pinus rigida, Robinia pseudoacacia, and Larix kaempferi35.

The vegetation climate zones of the Korean Peninsula are intricately linked to its climatic diversity and topographical variation. From north to south, the region transitions from deciduous forests characterized by broadleaf trees in the cooler, central zones to evergreen subtropical forests in the warmer, southern regions and on Jeju Island36. These zones are further delineated by elevation, with alpine vegetation prevailing at the highest altitudes, including the Seorak and Sobaek mountains. Each zone harbors a unique array of plant species adapted to specific climate conditions, contributing to the peninsula’s ecological richness37.

The variation in elevation from sea level to 2,744 m has a profound impact on flora distribution, with endemic plants showing a pronounced affinity for higher elevations. These areas, along with the rias coastal areas on the south and west coast and the islands isolated from the peninsula, are recognized as biodiversity hotspots, crucial for the conservation of unique plant species. However, the looming threat of climate change raises concerns about shifting distribution patterns and the potential migration of species to higher altitudes or latitudes in search of suitable habitats38.

Given its diverse topography, climate, and rich geological history, the Korean Peninsula serves as a natural laboratory for investigating the impacts of climate change on endemic plant species and vegetation climate zones. Its geographical isolation, enhanced by surrounding mountains and seas, introduces unique challenges in species dispersal and migration. The presence of distinct plant communities, adapted to the peninsula’s varied ecological zones, underscores the critical importance of studying both southern and northern lineage species for endemic plant, each exhibiting specific ecological preferences and adaptations39.

Species data collection

Data collection involved gathering coordinate information for 51,615 data points representing 360 endemic plant species, as designated by the Korea National Arboretum3. These data were obtained from ecosystem surveys conducted by various Korean government agencies, including the National Arboretum, National Institute of Ecology, National Forestry Research Institute, and National Institute of Environmental Science. The data were extracted from the National Herbarium (KH) database at the Korea National Arboretum, which holds specimens with GPS coordinates or generated coordinates from records, collected in South Korea between 1977 and 2022.

The coordinate data used in the modeling were derived from 309,333 specimens collected between 2003 and 2015, representing 2954 species across 175 families and 919 genera, to match the current simulation period37. The spatial error between plant growth points and GPS records was assessed by the National Arboretum to be within a maximum of 300 m, considering the typical survey recording practices and the technical positioning error of mobile GPS, which was kept within 10 m40.

Each survey was conducted by trained researchers following standardized protocols for species identification, including cross-referencing specimens where necessary to ensure reliability40. Additionally, since 2012, KH has been updating taxonomic positions, including scientific names, based on reports from the Korean Plant Names Index Committee to reflect changes in taxonomy41.

From this comprehensive dataset, we selected 179 species, each represented by at least 20 data points, based on their suitability for spatial statistical analysis; this selection criterion was designed to increase the robustness and representativeness of the study (Supplementary Table 1). Data processing included strict quality control measures, including coordinate accuracy verification and exclusion of duplicate records. The processed data were then spatialized as point data in vector format in the GRS80 ellipsoid-based Transverse Mercator (TM) coordinate system using QGIS 3.28 to identify patterns and correlations in species distributions.

Environmental data collection and processing

To examine the environmental influences on the habitats of our selected species, we reclassified environmental variables at a spatial resolution of 250 m, narrowing down to 38 critical variables. This refined set included 27 bioclimatic variables, 7 soil variables, and 4 topographic variables, as detailed in Supplementary Table 2. These variables were chosen based on their ecological relevance, supported by a thorough literature review42,43,44, and expert consultation with ecologists and biogeographers to ensure their appropriateness for our study region. The bioclimatic variables were derived from widely used climatic data sources to capture the climate-related aspects influencing species distribution, while soil and topographic variables were obtained from established soil databases and digital elevation models, respectively.

Exploratory data analysis (EDA) was conducted to ensure the completeness and accuracy of the spatial information. This process involved verifying the accuracy and consistency of the collected data, summarizing descriptive statistics, and identifying and addressing any missing values. This step ensured the dataset was robust and suitable for subsequent analyses.

To simplify the dataset and improve the analysis process, we used principal component analysis (PCA). PCA is a statistical method that helps address multicollinearity by transforming the original variables into a set of uncorrelated components, which prevents skewed results due to closely related variables. PCA has been shown to be effective in SDMs to reduce multicollinearity, as highlighted in recent studies45,46. We applied PCA to derive six principal components that collectively explained more than 90% of the variance in the observed environmental variables. These principal components were adopted as the final environmental variables for the study, enhancing the focus and efficiency of the analysis.

Species distribution model construction

In this research, we devised a systematic framework to develop species distribution models (SDMs) by correlating species occurrence records with environmental variables. This approach facilitated the creation of environmental profiles, enabling projections over different temporal and spatial scales. Utilizing the ‘biomod2’ package47 within R 4.3.0, we crafted an ensemble SDM employing a multi-algorithmic strategy for the analysis.

We chose an ensemble model to leverage the strengths of multiple algorithms and mitigate their individual weaknesses. Ensemble methods improve predictive performance and robustness by combining the outputs of various models, thereby providing a more reliable and accurate SDM48. This approach helps to reduce the risk of overfitting and increases the generalizability of the model across different temporal and spatial scales.

Our ensemble model integrated six statistical algorithms: Random Forest49, Generalized Additive Model50, Generalized Boosting Model51, Classification Tree Analysis52, Artificial Neural Network53, and Multiple Adaptive Regression Splines54. These algorithms were selected for their complementary strengths and widespread use in ecological modeling.

The inclusion of algorithms in species-specific models was contingent upon their predictive prowess, gauged by the Receiver Operating Characteristic (ROC) statistic, with a benchmark of 0.7 or above for algorithm selection23. This threshold guaranteed the incorporation of algorithms distinguished by their predictive accuracy, thus bolstering the ensemble model’s reliability and efficacy (Supplementary Table 3).To address the dataset composed solely of presence coordinates, we generated 1,000 random pseudo-absence points per model to achieve a balanced dataset. This process was confined to the South Korean territory to match the geographical range of the presence data. This ensures that the pseudo-absence points are representative of the area where the species may potentially be absent, thereby enhancing the robustness of the model. This approach is based on standard practices in SDM, particularly when dealing with presence-only data. By generating pseudo-absence points, we aim to mitigate the bias that can arise from using presence-only data, thereby enhancing the robustness of the model. This method is widely recognized and utilized in ecological modeling studies55,56. To generate the pseudo-absence points, we used the ‘biomod2’ package47 within R 4.3.0., which provides functions to randomly generate pseudo-absence points within the study area. The culmination of this process was the application of these meticulously constructed ensemble models to predict the potential distribution of each species across various current and future climate scenarios.

Analyzing geographic species range changes

Our methodology for assessing changes in geographic species range commenced with the transformation of ensemble probability maps into binary presence/absence maps, adhering to the True Skill Statistics (TSS) threshold criteria. This critical step allowed us to classify grid cells into categories representing suitable (1) and unsuitable (0) environments for species habitation, thus effectively delineating habitat suitability across the landscape.

To gauge the vulnerability of species to climate change, we evaluated their distribution changes over time under diverse scenarios against their existing distribution. This evaluation was quantified employing the Rate of Change Index (RCI) formula:

$$\:RCI=\frac{\left({Area}_{future}-{Area}_{current}\right)}{{Area}_{current}}\:\times\:100$$
(1)

Here, RCI symbolizes the percentage change in climatically suitable habitat range, with Areafuture and Areacurrent representing the future and current areas of suitable habitats, respectively.

Furthermore, to analyze the impact of climate change on the spatial distribution of endemic plant diversity, we compiled a species diversity map. This map was generated by layering the binary presence/absence maps of individual species across different timelines and climate scenarios. From these composite maps, we extracted the current state of species diversity and normalized species richness at each temporal snapshot to a scale from 0 to 1, using the formula ​11,23:

$$\:{SR}_{t}=\frac{{SC}_{t}}{{MaxSC}_{t}}$$
(2)

In this equation, SRt denotes the standardized species richness at time, SCt reflects the species count in each cell, and MaxSCt is the peak species count observed in each cell at time t. This normalization facilitates spatial comparison, especially in large regions, by providing an easier understanding of patterns in species richness​23.

Moreover, we estimated temporal differences in species richness(ΔSR) by subtracting the current SR from the projected SR in the last period(2100), for every grid cell ​23:

$$\:\varDelta\:SR={SR}_{2100}-{SR}_{\text{c}urrent}\:\:\:\:\:\:\:\:\:$$
(3)

The change (ΔSR) ranges between − 1 and 1. For grids with no species, the change was considered zero, assuming no variation occurred​23. This methodology enables a comprehensive analysis of the anticipated alterations in species distributions and diversity under varying climate change forecasts, facilitating a deeper insight into the potential ecological effects11,23.

We calculated the average longitude, latitude, and elevation of suitable habitats for each species simulated through species distribution models, both for the current distribution and under each climate scenario. Subsequently, we averaged these values across species to obtain data representing overall distributional changes by scenario and time period. This approach allowed us to analyze the distributional shift trends of endemic plant species under the scenarios.

Ecological niche overlap analysis

Niche overlap analyses using occurrence and environmental data can capture the likelihood of species interactions and potential changes due to range shifts57. By utilizing binary maps produced by species distribution models (SDMs), we converted suitable spaces into environmental space (e-space) by sampling conditions from suitable and unsuitable areas58. The e-space was then used to calculate niche overlap between species for the current period and future period (2071–2100). To measure niche overlap, we applied Schoener’s D59, which is calculated using the following formula:

$$\:D\left({p}_{X},\:{p}_{Y}\right)=1-\frac{1}{2}\sum\:\mid\:{p}_{X,i}-{p}_{Y,\:i}\mid\:$$
(4)

Here, PX, i and PY, i represent the frequency of species X and Y in the ith category, analogous to prey sizes. In the context of SDMs, these values indicate the relative suitability of habitats along an environmental gradient.

We classified endemic plant species into ‘Gain’ or ‘Loss’ categories by comparing the number of suitable habitat cells under current and future scenarios. ‘Gain’ species were those with increased suitable habitats in the future, while ‘Loss’ species had fewer. Following this classification, we calculated Schoener’s D for all species pairs for both the current and future periods (2071–2100) under each climate scenario to assess potential changes in niche overlap. Schoener’s D, which ranges from 0 (no overlap) to 1 (complete overlap), indicates the extent of environmental preference similarity between species. For instance, a decrease in Schoener’s D from 0.3 to 0.1 between Stewartia koreana and Abies koreana under the SSP5-8.5 scenario signifies a reduced probability of interaction, highlighting a potential shift in species dynamics. This detailed analysis showed that climate-driven habitat shifts could significantly alter species interactions and community composition, with important implications for ecosystem stability and biodiversity conservation. These calculations were carried out using the ‘Humboldt’ package in R 4.3.060.

Results

Changes in abundance and distribution of plant species under climate change scenarios

The analysis of current climate data revealed a pronounced abundance of endemic plant species within high-altitude areas above about 1000 m above sea level, notably around the Baekdu-daegan mountain range and along the southern and eastern coastlines near Jeju Island (Fig. 2, Supplementary Fig. 1). In these high-altitude regions, species richness tended to increase under various projected climate change scenarios. However, significant changes were observed in the low-altitude inland regions of South Korea, specifically in areas below approximately 300 m above sea level, where species richness tended to decrease with increasing climate scenario intensity. Additionally, there was a notable trend of species range expansion from South Korea into North Korea, which strengthened with the intensity of the climate scenarios (Fig. 2).

Fig. 2
figure 2

Projected changes in endemic plant species richness under SSP scenarios. This figure provides a comparative visualization of normalized species richness (SR) for 179 endemic plant species across different SSP scenarios over time. The SR scale, ranging from 0 (minimum) to 1 (maximum), indicates the normalized abundance of species, with delta SR (ΔSR) showcasing the net change between the current state and the scenario’s end (2071–2100). The maps were generated using ArcGIS Pro version 3.2 (Esri Inc., https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview).

Full size image

Characteristics of geographic distribution shifts under future climate scenarios

Our analysis identified expected shifts in the geographic distribution of endemic plant species under various future climate scenarios, as illustrated in Fig. 2. Specifically, endemic species are projected to adjust significantly towards higher altitudes and latitudes across all examined scenarios, suggesting an adaptation move towards elevated (Fig. 3-a) and more cooler habitats (Fig. 3-b). However, under the SSP3-7.0 scenario, the longitudinal shifts in species distribution show a significant increase from the current period until 2040, indicating notable changes in the east-west distribution. However, beyond 2040 to 2070 and into the 2071–2100 period, the distribution stabilizes, maintaining a pattern similar to the present configuration. This suggests that after an initial period of adjustment, the longitudinal distribution reaches a state of relative stability (Fig. 3-c).

Fig. 3
figure 3

Distributional shifts in plant species by elevation, latitude, and longitude under SSP scenarios. The figures show shifts in (a) elevation, (b) latitude, and (c) longitude for endemic plant species under SSP1-2.6, SSP3-7.0, and SSP5-8.5 scenarios across four time periods (current, 2011–2040, 2041–2070, and 2071–2100). Each point represents the mean value, with error bars indicating the standard error.

Full size image

Species-specific climate adaptations and area changes

The vulnerability assessment, based on warming intensity across scenarios, showed a direct correlation between the degree of warming and the likelihood of habitat change (Fig. 4, Supplementary Fig. 2). As the intensity of the warming scenarios increased, heightened sensitivity to habitat changes was observed in more species, resulting in a reduction of their habitat areas compared to the present day.

Fig. 4
figure 4

Sensitivity and Distribution Area Changes of Plant Groups under SSP Scenarios (2070–2100). Panels display the sensitivity (percentage, left) and change in distribution area (square kilometers, right) for endemic plant groups under SSP1-2.6, SSP3-7.0, and SSP5-8.5 scenarios. The violin plots illustrate the variability and distribution of sensitivity and area changes across these scenarios during the simulation period from 2070 to 2100. For graphs detailing sensitivity and area changes for individual species, refer to Supplementary Fig. 2.

Full size image

Table 1 illuminates these trends further by indicating the proportions of species experiencing changes in their distribution. Under the SSP1-2.6 scenario, 39.1% of endemic herbs are projected to expand their range, suggesting a more favorable outcome under moderate climate change, while 35.2% might see a decrease. Shrubs and trees also show potential for expansion, with 9.5% of shrubs and 5% of trees expected to increase their range. As climate change intensity increases under the SSP3-7.0 scenario, the percentage of herbs likely to decrease in range rises to 40.8%, with only 33.5% experiencing an increase. Similarly, 10.1% of shrubs are expected to decrease, while only 6.7% might increase, and 4.5% of trees may expand compared to 2% decreasing. In the SSP5-8.5 scenario, representing the most severe climate change, 41.9% of herbs are expected to experience range contraction and only 32.4% expanding, with shrubs showing 11.2% decreasing and only 5.6% increasing, and trees remaining stable with 4.5% expansion against 2.2% contraction. Overall, the data indicates that harsher climate conditions lead to greater range decreases, particularly for herbs, while trees and shrubs experience less dramatic changes but still reflect a trend towards increased range loss as climate change intensifies.

Table 1 Proportional changes in endemic species distribution across vegetation types relative to total species under SSP scenarios (2071–2100).
Full size table

Stability and variation in ecological niche overlap under climate change

The pattern of niche overlap among all endemic plant species was consistent across all scenarios, with the average degree of overlap showing little change, suggesting that ecological niches remained relatively stable despite range shifts due to changing climate conditions. While species with shrinking ranges experienced a slight decrease in average overlap strength, it did not show significant changes. In contrast, species with expanding ranges exhibited a notable increase in overlap strength compared to the other two groups (entire group and shrinking range species), indicating a tendency for greater overlap under all climate scenarios. Despite the overall average overlap remaining stable, the overlap strength between individual species within each group varied significantly, reflecting substantial changes in interactions at the species level (Fig. 5, Supplementary Figs. 3, 4).

Fig. 5
figure 5

Figures detail changes in niche overlap (Schoener’s D) for plant groups, during current times and the SSP 3–7.0 scenario (2071–2100). Grey lines represent species overlap intensity, with dotted black lines connecting group median values. ‘Gain’ refers to the group of species for which there are more suitable cells under future scenarios than in the current habitat, indicating area expansion. ‘Loss’ indicates the opposite group, where there are fewer suitable cells in future scenarios, signifying area reduction for endemic plants. See Supplementary Fig. 4 for graphs for other scenario.

Full size image

Discussions

In the face of accelerating climate change, understanding the distribution dynamics of plant species within terrestrial ecosystems has never been more crucial. The intricate web of habitat overlaps, and interspecies interactions forms the backbone of biodiversity, influencing ecological stability and resilience61,62. This study’s exploration into the spatial-based predictions of species composition changes due to climate change in Korea represents a significant advancement in ecological modeling, providing view of potential future scenarios for endemic plants.

While our study primarily focuses on the endemic plants of South Korea due to limited access to reliable data from North Korea, it is important to recognize the inherent limitations this poses. The data available from platforms like GBIF for North Korean species is often unverified and may not accurately reflect their original habitats, thus complicating the modeling of their distributions. Despite these challenges, we included the entire Korean Peninsula in our analysis to account for the anticipated northward shift in species distributions driven by climate change29. This approach, although constrained by data availability, provides a more holistic and scientifically robust projection of future distribution scenarios63,64.

There are well-documented limitations of using biased occurrence data in species distribution models65,66,67,68. Despite these limitations, our inclusion of North Korean regions in the models underscores the interconnected nature of the peninsula’s ecosystems and highlights the necessity of comprehensive cross-border environmental studies. By addressing these limitations transparently, we aim to encourage future research efforts that might bridge these data gaps and enhance the accuracy of species distribution models in this region. This holistic approach remains a reasonable and widely accepted method in the field69,70,71, ensuring that our findings contribute valuable insights into the impact of climate change on endemic plant species. Notably, the modeling algorithm employed here, like most within the domain of species distribution modeling, is constrained by its limited ability to factor in life history elements such as species dispersal ability and survival rates. Despite these limitations, our methodology distinguishes itself by incorporating a wider array of variables, including those altered over geological timescales—soil and topography—besides the bioclimatic variables typically considered. This inclusivity enhances the realism of our species establishment predictions, providing a more comprehensive understanding of species resilience and vulnerability under changing climatic conditions65.

Our predictions indicate a general trend of endemic plants moving toward higher latitudes and elevations, suggesting increased species overlap in these regions72. Currently, high-altitude forest areas such as the Baekdu-daegan mountain range exhibit a high abundance of endemic plants and are identified as hotspots under all climate change scenarios (see Supplementary Fig. 1). These endemic plants, with their restricted distribution ranges, specialized environmental conditions, and small populations, are considered remnants of past climatic changes73. Historically, these plants have been confined to high-elevation areas, serving as climatic refuges during interglacial warming periods74,75. However, as environmental conditions shift, the species using these refuges may change. For example, the isolated geographical nature of Mt. Halla on Jeju Island, another current hotspot for endemic species, restricts the dispersal abilities of these plants, potentially leading to their extinction as climate change progresses. If they cannot adapt or migrate to new suitable areas, they will likely disappear and be replaced by species better adapted to new conditions76. This shift will alter species diversity and composition in these regions. Similar risks are observed in studies examining terrestrial vertebrates in North America, underscoring the importance of dispersal for maintaining biodiversity in climate refuges11.

The unique topography of the Korean Peninsula, formed since the last glacial period, amplifies the anticipated impacts of climate change on the spatial overlap of plant species77. However, the current distribution of these endemic species cannot be attributed solely to natural selection, as areas with low land use intensity due to minimal anthropogenic pressure have also significantly influenced their habitats. This underscores the potential role of conservation efforts in expanding and maintaining the habitats of endemic species beyond their current boundaries. Examples of such conservation efforts include habitat restoration78 or in situ conservation79, the establishment of protected areas80, and ex situ conservation through assisted migration81. Conversely, lowland areas such as the inland regions of South Korea are subject to intensive land use, limiting the introduction of new species and further restricting the habitats of endemic species. This highlights the differential impacts of climate change on biodiversity across geographical areas, suggesting spatial variability in biodiversity changes across the Korean Peninsula. Therefore, tailored conservation strategies that consider these regional characteristics are necessary. Efforts to enhance habitat connectivity and support species movement and adaptation should reflect the unique ecological, geographical conditions, and land use contexts of each region. Such strategies will enhance biodiversity resilience to climate change and enable sustainable conservation management.

Our study highlights the potential shifts in distribution and ecological niches of endemic plant species under climate change, laying a foundation for future research and conservation planning. The comprehensive database we developed provides valuable data for crafting conservation strategies that address the specific vulnerabilities of plant groups, with a focus on active measures like species relocation and habitat creation to mitigate climate impacts. The findings emphasize the need for policy frameworks that incorporate these strategies and adapt to regional ecological and geographical contexts. Aligning conservation efforts with the specific needs of endemic species and promoting sustainable land use practices can enhance biodiversity resilience and support long-term ecosystem stability across the Korean Peninsula. Additionally, public engagement and awareness campaigns should be designed to complement these policies, ensuring broader societal support for conservation initiatives. By integrating predictive modeling with ecological theory, this research deepens our understanding of plant distributions in Korea and guides effective ecosystem management in response to climate-driven changes. Continued research will be essential to address the challenges and opportunities presented by these ecological shifts.

Conclusions

This study has undertaken a comprehensive analysis to predict the short- and long-term distribution changes of endemic plants on the Korean Peninsula under various future climate change scenarios. The findings reveal significant insights into the adaptive responses of these plant species, the implications for biodiversity, and the considerations necessary for effective conservation policy.

  1. 1

    We found that endemic plant species exhibit different degrees of distribution change depending on the climate scenario, influenced by current habitat characteristics, range size and degree of habitat dispersal. Significant differences in vulnerability were found between southern and northern groups of endemic species, highlighting the need for further research using different criteria to categories endemic species groups. This approach will improve our understanding of changes in ecosystem processes.

  2. 2.

    Predictions indicate that under different climate scenarios, the habitats of endemic plants will shift to higher altitudes and latitudes. However, it should be noted that the simulation area of this study is the entire Korean Peninsula, while the data used to build the model were limited to Korea due to limited availability, so the degree of migration to higher latitudes may be somewhat overestimated.

  3. 3.

    Currently, areas of high endemic plant abundance include the high mountains around Baekdu-daegan and the coastal islands. Climate scenarios suggest a macroscopic shift of these high-abundance areas to higher altitudes and latitudes. However, within these broad shifts, there may be variation due to differences in species composition.

  4. 4.

    The study expects different extents and intensities of interspecific overlap in different areas, driven by the different ranges of each species. Endemic plants, characterized by their limited range, are expected to show less intra-specific overlap and variation compared to the plant communities occupying similar environmental ranges.

  5. 5.

    The results of this research provide a fundamental basis for formulating conservation policies for endemic plant species and for guiding various subsequent studies. Specifically, we suggest further investigations to compare response trends of different functional groups and to predict changes in pollinator nesting patterns. Such studies are essential to elucidate the ecological characteristics of endemic plant species and to assess potential shifts in ecosystem function.