Introduction

Landscape composition, configuration, and topography are essential factors that determine bird distribution patterns. These factors are crucial for anticipating potential impacts of land use changes and formulating future conservation strategies. Concurrently, recognizing the fundamental interconnections between land use and biodiversity is significant for illuminating the intricate relationships between human societies and their surrounding environments.

Throughout history, extensive changes to the Earth’s land surface from land use activities—including the conversion of natural landscapes for human purposes and modifications in management practices on human-dominated lands—have significantly transformed these areas. Such alterations are widely believed to contribute to biodiversity declines, manifested through habitat loss, modification, and fragmentation, degradation of soil and water, and overexploitation of indigenous species1. Conversely, land use diversity, which denotes the variety of land use categories within a specific area, is often regarded as a vital environmental factor that fosters species richness at both landscape and regional scales by enhancing beta-diversity2,3.

Biological diversity metrics, acquired through surveys, are helpful indices to quantify species-level biodiversity and often represent the ecosystem’s status and integrity. Taxonomic diversity, defined by species richness and abundance, is traditionally and widely used to explore one of the characteristics of ecological communities. Functional diversity focuses on more systematic aspects of biodiversity interacting with the ecosystem processes4. It considers functional traits—such as morphological, structural, biochemical, physical, or behavioral characteristics of a species—which delineate a species’ ecological niche and are increasingly adopted to understand the linkages between species and ecosystem functioning4.

By integrating diversity metrics that prioritize functional attributes, this approach provides a comprehensive framework that enhances the conventional taxonomic perspective. Functional diversity can capture the array of functional roles performed by species within a community, while species traits specifically detail the influence of these species on ecological processes within a particular assemblage, thus encompassing broader ecological phenomena5,6.

Understanding the spatial variation in taxonomic and functional diversity allows exploring broader biotic and abiotic interactions. Developing geospatial tools in conjunction with geospatial data detailing species distribution across large spatial extents has facilitated the exploration of relationships between biodiversity and landscape composition and configuration. As land use and land cover changes often lead to changes in primary and secondary producers in an ecosystem and subsequent food resources and barriers or threats for animals, wildlife assemblages and biodiversity are influenced by landscape changes7. Although tendencies can vary depending on scale, as the intermediate disturbance hypothesis supports2, areas with heterogeneous landscapes highly contribute to wildlife biodiversity in general8,9. Understanding relationships between biodiversity and landscape structure helps find ecological patterns at a large scale and informs conservation planning5.

Aside from the regional importance as a habitat for dozens of resident birds, the Korean Peninsula serves as a seminal migratory pathway within the East Asia-Pacific migratory route, which is recognized as one of the principal global flyways for birds10,11. Various topographic features on the Korean Peninsula, including lakes, rivers, estuaries, mountains, hills, and vast areas of rice fields and tidal flats, provide various kinds of habitats for avian species with different ecological niches. Avian fauna in South Korea has been investigated through national environmental surveys since 1986 by the Ministry of Environment, confirming regional and national distributions of birds and their abundance.

This study investigates the interplay between landscape features and avian diversity in South Korea, as well as the correlation between taxonomic diversity and functional diversity, utilizing an extensive dataset collected at a national level. Following the intermediate disturbance hypothesis2, we predict that taxonomic and functional diversity levels will be highest in open forests or forest edges and lowest in agricultural or urban areas with intensive use.

Understanding the overarching role of land use and land use diversity in shaping regional diversity across the country is paramount in evaluating the significance of large-scale environmental complexity beyond the quality of individual land use categories8. Moreover, it aids in developing policies aimed at mitigating biodiversity loss at the regional scale. As the number of birds found in South Korea has been increasing in recent decades, including more frequent visits by subtropical and tropical birds, and the distributions of many birds have changed due to climate change and urbanization, this study provides baseline information about avian diversity distribution on a national scale at a critical moment.

Results

Assessing bird-environment relationships through biodiversity metrics

Various biodiversity indices displayed different aspects of bird diversity in South Korea (Fig. 1, Supplementary Figure 1). Taxonomic diversity of birds found in South Korea was not highly related to functional diversity indices (Fig. 1, Supplementary Figure 2). In general, of the bird taxonomic diversity indices, Shannon diversity index was noticeably high in the vast plains in the southwestern part of South Korea (Fig. 1A,B). While being evenly distributed over the country, species richness (SR) tends to show higher richness on relatively lower elevations and plains. On the other hand, functional richness (FRic) and functional divergence (FDiv) were higher in shorelines, whereas functional evenness (FEve) was higher in mountainous areas (Fig. 1C,D,E). Bare/sparse vegetation, herbaceous wetland and permanent waterbody were not clearly relevant to bird diversity indices (Fig. 2, MCMCglmm p > 0.05).

Fig. 1
figure 1

Diversity indices of birds mapped across South Korea. (A) Shannon diversity index (H), (B) Species richness (SR), (C) Functional richness (FRic), (D) Functional evenness (FEve), (E) Functional divergence (FDiv). The output maps were categorized into 6 classes using the Jenks natural breaks method. ‘NA’ indicates sites where the survey was not conducted.

Full size image
Fig. 2
figure 2

Effect sizes of landscape features on biodiversity indices. (A) Posterior means and 95% credible intervals from MCMCglmm for each variable of the landscape elements. Asterisks denote significance level for the pMCMC statistic, with * representing p < 0.05. (B–F) Scatter plots displaying only significantly different categories indicated in (A), with the black line depicting the trend. The y-axis represents diversity indices, while the x-axis represents environmental conditions. Both axes range from 0 to 1. SR denotes species richness; H denotes Shannon index of diversity; FRic denotes functional richness; FEve denotes functional evenness; FDiv denotes functional divergence.

Full size image

Specifically, for SR, a taxonomic diversity index, more extensive areas of urbanized terrains and closed forests significantly lowered the bird diversity (MCMCglmm, p < 0.05). However, richly-shrubbed areas were positively associated with SR of birds (MCMCglmm, p < 0.05). Shannon diversity index (H) had a significant negative correlation with larger area of herbaceous vegetation (MCMCglmm, p < 0.05). The Shannon index increased at the coast as well as in areas with bigger open forests but was negatively related to human population size (MCMCglmm, p < 0.05). FRic, a functional diversity index, was significantly lower at lower elevations, larger urbanized terrains, larger closed forests, larger agricultural fields, and bigger human population size, while it was higher on land with heterogenous landscapes or at the shore (MCMCglmm, p < 0.05). FEve was significantly lower where the area of open forest was larger, slope was steeper, and land cover diversity was higher (MCMCglmm, p < 0.05). FDiv was significantly lower in urban areas, agricultural field, closed forest, open forest and steep inclined areas, while more diverse land types raised the FDiv value (MCMCglmm, p < 0.05).

Discussion

Recent studies show a positive correlation between land use diversity and both species richness and functional richness, suggesting that diverse land use can enhance ecological diversity8,12. However, this study, which examined land use diversity across the entirety of South Korea, revealed contrasting results, as we found no significant relationship between land use diversity and species richness. The contrast may be attributed to differences in the scale of analysis. Previous studies focused on broader geographic areas with varying climate zones, which can influence bird assemblages and communities differently than the relatively homogeneous conditions found in South Korea8. Additionally, the impact of human activities on land use and biodiversity may vary significantly across different climates, affecting the generalizability of previous findings13,14.

Looking closer to South Korea on a national scale, we found FRic and FDiv higher when the land cover was more diverse, while FEve was lower. Land use diversity was higher in urban or populated areas with urban infrastructure and some natural features. In contrast, deep forests in mountainous regions in the eastern part of South Korea had lower land cover diversity. Interestingly, diversity trends in closed forests were in line with those in urban/built-up areas, although urban areas exhibited greater diversity of land use and land cover than densely forested mountainous regions. Birds in closed forests were lower in taxonomic and functional diversity. These trends were also found in agricultural fields with intense farming, allowing only a few species to inhabit the simplified environment15.

Open forests and closed forests exhibit markedly different trends in avian diversity. Open forests or forest edges often share characteristics with transition zones to some extent, while closed forests typically represent relatively monotonous and uniform landscape compositions16,17. Transition zones between two land types are likely to support higher biodiversity, including both functional and taxonomic diversity, as species from distinct zones can intermingle in these regions. Landscapes in open forests tend to incorporate nearby landscape types, leading to higher diversity of land cover and species. However, this does not imply that thick forests are less valuable for birds than open forests. Closed forests provide essential habitats for particular species, such as Halcyon coromanda, which require shaded environments and fewer disturbances11,18.

Urban areas, despite often featuring low biodiversity similar to that in closed forested regions, contain a variety of landscape elements such as parks, street trees, fountains, sewage systems, and even cracks in cement walls, which can support specific flora and fauna. Studies have shown that higher landscape diversity in urban settings leads to increased biodiversity17,19. This effect is especially pronounced on a local scale, where detailed examination reveals significant diversity within specific urban areas. However, on a larger scale, such as the one examined in our study, urban areas represented by aggregated plots of densely populated regions exhibit lower avian diversity despite the presence of diverse landscape elements. This suggests that large, consolidated urbanized areas tend to support less avian diversity compared to other land use types with varied landscape features.

In South Korea, urban environments are often characterized by highly dense, built-up areas that limit species diversity. Our study supports the finding that urbanization negatively impacts bird diversity, showing a significant decline in avian populations in urban areas compared to non-urbanized settings20,21,22. Nevertheless, as many studies have revealed, urban areas have the potential to secure higher biodiversity through thoughtful urban planning and management strategies17,23. On a broader scale, it is also essential to consider the connectivity of ecological networks and systems across neighboring cities.

Species richness was evenly spread across the country, but according to the Shannon index, taking into account species richness and evenness, areas characterized by less urbanized open plains exhibit higher taxonomic diversity. On the other hand, coastal regions had higher functional indices, FRic and FDiv, implying that a diverse array of water birds with distinct traits were attracted to these areas. Many different niches and drastically divergent traits of birds might have been satisfied in coastal areas, underscoring that land use and land cover changes in the region might be much more influential on bird diversity. Particularly, vast tidal flats on the west and south of the Korean Peninsula have been essential habitats for many fish, invertebrates, and birds, including migratory and residents24. Migratory birds occupying niches while moving with seasonal changes might have contributed to higher functional diversity.

Lower FEve along the shorelines implied that birds with certain traits were much more abundant than others. The lower FEve on the coast can be explained by the flock size that water birds formed. In the Korean Peninsula, migratory water birds such as ducks, gulls, and sandpipers often form larger groups of several hundred to a thousand individuals, whereas predatory water birds such as osprey and white-tailed eagle barely flock together, although they may gather in large numbers during migration25,26. On the other hand, forest birds rarely form flocks larger than tens to a hundred individuals, limiting difference in evenness measures5,25. Higher FEve and lower FRic and FDiv in the deep mountain forests on the Peninsula suggested that functional distributions of forest birds were similar in number. In contrast, the bird species found in the forests were not particularly functionally diverse.

Notably, these bird diversity indices consistently displayed stable patterns corresponding to different land use types. Consequently, the findings allowed us to make accurate predictions regarding bird diversity indices based on the specific land use and land cover configurations. Moreover, the observed complementary relations among diverse biodiversity indices indicated that each index provided distinct and valuable insights into the overall biodiversity of the region. Considering the overarching tendency of bird species to inhabit specific regions, our study concludes that functional biodiversity indices complement taxonomic diversity indices in elucidating bird distribution patterns.

This study is significant as it provides a comprehensive analysis of the functional diversity of birds across various landscape units at a national scale. Calculating functional diversity at larger scales poses challenges due to the complexities involved in accurately measuring abundance and individual numbers. By employing a regularized grid system to represent landscape variability in South Korea, we were able to generalize patterns and trends effectively.

Our findings highlight the importance of considering both regional and national scales when assessing biodiversity. The comprehensive inventory and diversity assessment of regional bird species, when integrated with our data, offer valuable insights into the intricate relationships between species and their landscapes. These insights are crucial for informing land planning strategies that prioritize biodiversity conservation.

Additionally, our study underscores the need for large-scale ecological studies to incorporate functional diversity metrics, which provide a more nuanced understanding of ecosystem status and species interactions. The ability to generalize findings across extensive geographic areas can significantly enhance our approach to biodiversity conservation, ensuring that strategies are both effective and sustainable.

Methods

Species abundance and occurrence data collection

Bird data were referenced from the third, and fourth national investigations of the natural environment (Chŏn’gukchayŏnhwan’gyŏngjosa, hereafter NINE) conducted by the Ministry of Environment, Republic of Korea. The third NINE was conducted from 2006 to 201227, and the fourth from 2014 to 201828 (The data can be found at: https://www.nie-ecobank.kr/).

For the NINE survey, main survey sites were determined using a 1:25,000 topographical map created by the National Geographic Information Institute. Based on grid compartments, each site was segmented into nine unit survey plots, with each plot measuring 2′30″ × 2′30″ in DMS (degrees, minutes, seconds) notation. The area encompassed by 2′30″ × 2′30″ block is estimated to be approximately 13.65 square kilometers, with each side ranging from approximately 3.5 to 4.5 kilometers. It is worth noting that this calculation can vary depending on the specific latitude of the block. Consequently, a total of 6,748 plots (grid cells), each corresponding to these angular measurements, were established across South Korea (see Fig. 3). Bird surveys were conducted by visual and auditory observations along designated line transects. The survey period spanned from March to November, targeting winter (March/November) bird populations, breeding birds (April–June), and migrating birds (April–May/September–October), with each season surveyed once. The data from each plot were organized into arrays, taking into account seasonal variations in the species abundance, and mean values were subsequently calculated for taxonomic and functional diversity. For each survey plot, the mean values obtained by survey period were derived, and the mean values of nine survey plots were entered as the value of one site (grid). Bayesian statistics were then applied for further analysis. Additionally, we provide the “DAVIT_SpLocation” to document where each species was observed within each plot.

Fig. 3
figure 3

Grid system for surveying biological species in South Korea.

Full size image

The NINE provides a brief survey report, and records including species abundance data with GPS coordinates are appended to the report29. We mapped the point data using QGIS (ver. 3.30.2) and verified whether the species data points were located in plausible areas based on the expertise. When a GPS error was found, the location was corrected accordingly by checking the identification number of plot assigned on each grid and converting the exact address on the report to the GPS point. If GPS data were missing for a species report, the original report was checked for the grid numbers to assign species presence information. As a result of compiling four surveys, 385 species of birds and a total of 5,855,184 abundance birds were presented with a total distribution data of 745,943 data (“K_DAVIT_ObSpList” in Figshare30). Nomenclature follows the National List of Species of Korea (NLSK) provided by the National Institute of Biological Resources31.

To assess the potential sampling bias inherent in the raw data, we utilized the R package sampbias (ver. 2.0.0)32, adhering to the guidelines provided by the developer. Our selection of candidate biasing factors encompassed roads, mountain trails, rivers, and cities, sourced from the National Spatial Data Infrastructure Portal (http://data.nsdi.go.kr/ Accessed on Jul 1 2022) and the public data portal (https://www.data.go.kr/ Accessed on Jul 1 2022), respectively. The raster resolution employed for calculating bias was set at 0.01 (refer to Supplementary Figure 3). Notably, the results derived from the sampbias analysis did not reveal any statistically significant sampling bias, leading us to omit detailed discussions within the Results section.

The singular biasing factor demonstrating a substantial posterior weight was ‘roads.’ However, the sampling rate dependent on ‘roads’ exhibited a consistent pattern, barring a few instances at specific distance points. Additionally, the projection of the estimated sampling rate by biasing factors demonstrated analogous values across South Korea. It is imperative to note that, despite the prominence of ‘roads’ as a biasing factor, our findings underscore the overall reliability and representativeness of the dataset, thus contributing to the robustness of our research outcomes.

Species trait data collection

The functional traits of 385 species were compiled according to the records on the NINE surveys (“K_DAVIT_TraitData” in Figshare30). Diet, foraging location, and morphological features were treated as continuous traits33, while habitat location and nesting location were categorized34 (Table 1). Diet consists of ‘invertebrates’, ‘vertebrates’, ‘herptiles’, ‘fish’, ‘decaying biomass’, ‘fruit’, ‘nectar’, ‘seed’, and ‘plant’. Foraging location consists of ‘below surface water’, ‘around surface water’, ‘ground’, ‘understory tree’, ‘mid-high tree’, ‘canopy’, ‘aerial’, ‘pelagic specialist’, and ‘nocturnal specialist’. ‘Pelagic specialist’ and ‘nocturnal specialist’ are marked with binary values. As for morphological features, average body mass is presented. Habitat type is categorized as ‘urban area’, ‘agricultural area’, ‘coast’, ‘lake-river’, ‘wetland’, ‘grassland and/or shrub’, ‘forest’, and ‘mountain and/or alpine’. Nesting locations are classified as ‘ground’, ‘grass and/or shrub’, ‘tree’, ‘tree hole’, ‘cliff and/or cave’, ‘water level or near water’, ‘artificial structure’, ‘parasitize’, and ‘reuse’. For analysis, the categorical data were converted into binary values, where each categorical data value was represented as either 1 (observed within the grid cell) or 0 (not observed).

Table 1 Bird functional traits used in the study.
Full size table

Prior to compiling traits and distribution data, species inventory was organized following the NLSK. Common names in Korean and scientific names of some species were matched based on recent scientific discoveries in the latest version of the IUCN Red List of Threatened Species (https://www.iucnredlist.org/ Accessed on Jul 1, 2022). Traits data of birds were referred to EltonTraits and JAVIAN database. Lacking traits information in the database of EltonTraits and JAVIAN were filled with the data from IUCN and research papers34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59.

Taxonomic and functional diversity calculations

Taxonomic diversity was represented with species richness and the Shannon Index. Functional diversity indices including functional richness (FRic), functional evenness (FEve), and functional divergence (FDiv) were calculated using the FD package (ver. 1.0) in R60. FRic, or functional richness, measures the size of the space occupied by the traits of species within an ecosystem. To calculate this, principal component analysis (PCA) is first performed on the trait data to capture the main variations. Subsequently, the minimum volume convex hull that includes all species is calculated based on the space defined by the extracted principal components. This process quantitatively represents how diverse the functional traits are across the species in the ecosystem61. FEve, or functional evenness, indicates how evenly species are distributed within a functional trait space in a community. It measures the evenness by calculating the relative distances between species in this space, considering their abundances. This index provides insights into whether some species dominate specific functional niches or if there is an even spread across the functional traits represented in the community61. FDiv, or functional divergence, measures the variance in the distribution of species within the functional trait space, weighted by species abundance. It describes how far and in what manner species diverge from the centroid of the functional space, emphasizing the roles of more abundant species in shaping the community’s functional structure61.

Environmental data collection

The relationship between biological diversity indices and environmental elements was explored through various factors, including land use and land cover (LULC), digital elevation model (DEM), slope, and human population. LULC was analyzed with Copernicus dataset with a resolution of 100 m62. We consolidated the subcategories of closed forests, including ‘evergreen needle leaves’, ‘evergreen broad leaves’, ‘deciduous needle leaves’, ‘deciduous broad leaves’, ‘mixed’, and ‘not matching’ into a unified category of closed forest. The subcategories under the open forests were also not specified but integrated as the “open forest”. The area of each land type in a grid was summed respectively, and LULC diversity in a site was measured using the Shannon index by putting area instead of species abundance. The calculated LULC diversity represented heterogeneity of the land; the higher the value indicates more diverse landscape patches (Supplementary Figure 4). The DEM and the slope were obtained from SRTM Digital Elevation 30 m with average values in each site. All data above were clipped and adjusted to the area of South Korea and downloaded through the google earth engine (https://hyunkim36.users.earthengine.app/view/k-davit-fd). Population data were built based on a national census collected by administrative district in 2021. Since the standard grid did not exactly match with existing administrative districts, the population at each site was estimated by summing the populations of the smallest administrative units.

Statistical analysis

We employed a Bayesian statistical approach using the MCMCglmm package version 2.3463 to examine the relationships between ecosystem diversity indices and environmental factors. Prior to the analyses, all data were rescaled between 0 and 1 to satisfy the assumptions required by the model. The analysis framework utilized ‘site’ as a random factor with a prior specified as a multivariate normal distribution, where both the mean vector and the diagonal covariance matrix were set to 0 and 1, respectively, and the degree of freedom parameter was set at 1. For fixed effects, the prior was a multivariate normal distribution with a zero mean and a diagonal covariance matrix scaled by 10,000, indicating a strong prior belief of no effect prior to observing the data. The residual variance was modeled assuming an inverse gamma distribution with a shape parameter of 0.002. The model ran for a total of 900,000 MCMC iterations, with the first 500,000 iterations discarded as burn-in and subsequent samples thinned every 500 iterations to reduce autocorrelation. We ensured the robustness of our analysis by conducting autocorrelation diagnostics on the posterior samples of both the fixed effect parameters and the variance-covariance matrix for the random effects, maintaining all autocorrelation values under 0.1. Correlations among indices and relevance within environmental factors were ascertained by Pearson correlation analysis using R Package corrr ver. 0.4.4.