Introduction

In an era of global environmental change, the persistence of populations in newly-unsuitable conditions and the establishment of new populations following disturbance are key in determining changes in biodiversity over time and in space. For vascular plants, the soil seed bank can play a vital role in both processes. Soil seed banks are the communities of persistent seeds being naturally stored in the soil that contribute to vegetation regeneration after disturbance1,2. Plant species that form seed banks do so as part of a life-history strategy that allows for the buffering of temporal environmental heterogeneity3–5, and as such, seed banks are an important component for understanding plant (meta)population and community dynamics6,7.

Despite the potential advantage of producing seeds that persist in the soil, not all species contribute to soil seed banks. Instead, the ability to form seed banks is a functional strategy for plants to (re)establish and maintain populations, which has evolved in response to both predictable and unpredictable disturbance regimes8. The species that do form seed banks also vary in terms of the number of seeds produced, and the longevity of seeds in the soil9. The disturbances that have contributed to the development of seed banking are not randomly distributed in space, and as such plant communities in different bioclimatic environments should rely on seed banking to different extents, with implications for local seed bank size and composition. For example, in arid systems dominated by drought, it is advantageous to have an annual life history and form persistent seed banks to buffer annual changes in water availability10,11. A similar pattern exists in temperate Europe, where species with warmer distributions have been found to have a higher seed bank persistence than more northerly species12. In tropical systems, where climatic conditions are broadly stable and exposure to freezing temperatures and drought conditions is rare, relatively few species produce seeds that are able to survive desiccation13,14—a trait associated with long-term persistence—which may result in relatively small and species-poor soil seed banks.

While it can be interesting to broadly compare plant communities from different world regions, ecological assemblages can differ considerably among habitats within the same region. The frequency of disturbance within a habitat is generally mirrored by the strategies of the resulting plant communities15. Therefore, different habitats could be expected to support different soil seed banks, in terms of the number of species and the density of seeds present. For example, in temperate and boreal regions, forests are dominated by long-lived trees, with the understorey often covered by species that largely rely on vegetative reproduction rather than seed banking16,17. Species-rich grazed grasslands in the same regions are characterised by regular, low-intensity and spatially-heterogeneous disturbances, and as such harbour plants with a wide range of regeneration strategies18,19. Intensively-used arable lands on the other hand are well-known to support large communities of ruderal species with annual life-histories and high seed production20,21. As such, habitat is an important component in understanding how different functional traits are represented in different regions, while differences between broad habitat types can differ in different world regions22. Moreover, communities in the same broad habitat categories can differ in their responses to environmental degradation in different regions23. Therefore, when studying plant communities at global scales, it is important to consider the potential roles of different habitats and environmental degradation driving any variation within or between regions. This could be especially relevant in the case of soil seed banks, which exist partially as a strategy to respond to disturbance.

Studying patterns of soil seed bank diversity across large spatial scales and geographic gradients can be useful for understanding the biogeography and macroecology of their role in biodiversity maintenance and recovery24–26. However, biodiversity is known to be scale-dependent, and comparisons of soil seed banks across ecoregions and habitats are therefore likely to be influenced by the scale at which diversity is estimated. It has been established that the number of species in soil seed banks increases with area24,25, but while it is well-known that this accumulation of species varies according to habitat and region in the established vegetation as well as in other taxa27–30, such patterns are still largely unknown in the soil seed bank, at least not at spatial scales that span multiple biomes. Integrating a scale-explicit approach to understanding diversity in the soil seed bank across habitats, within and between ecoregions and in response to habitat degradation is an important step for plant community and macroecology, allowing critical insights for both broad-scale patterns of biodiversity and the context-dependent interpretations of local dynamics.

While individual seed bank studies are usually (through necessity) small-scale comparisons across limited environmental gradients within a specific region, global analysis allows the large-scale comparison of biodiversity of the seed bank. Here, we use a global database of soil seed banks, harnessing 1442 studies from all seven of the world’s continents (Fig. 1, Table S1), to reveal how broad differences in climatic conditions across world regions, and differences in disturbance regimes across habitat types, combine to determine patterns in both the richness and density of soil seed bank communities at different spatial scales. The global seed bank database31 is the result of a comprehensive search of the seed bank literature, providing 3096 records containing information regarding the richness and/or density of the soil seed bank at a given location (‘Methods’). To investigate patterns of seed bank diversity among ecosystems and global biomes, we used both the WWF map of global ecoregions32 and the habitat descriptions used by the authors of the component studies to separate records into 15 ecosystems. These ecosystems are defined as the combination of biome (e.g. forest, grassland) and ecoregion (e.g. temperate, tropical). Each ecosystem was assigned to either the terrestrial, aquatic or transitional realm33; see Fig. 1 and Table 1.

Fig. 1: The global record of the soil seed bank.Points indicate the location of each record collected from the literature (n = 3096; some records have the same location). Records are first split according to biome (e.g. tundra, forests; a–g), with each biome split according to one or more ecoregions (e.g. boreal, tropical; colours of points). The combination of biome and ecoregion results in 15 unique ecosystems. Each record is also designated either as occurring in undisturbed or degraded ecosystems, apart from arable records which have their own category (shapes of points). We also categorise each biome as belonging to the a–e Terrestrial, f Transitional, or g Aquatic realm. See ‘Methods’ for a more detailed description of all categories. Background maps are public domain, see https://www.naturalearthdata.com.

Realm | Biome | Ecoregion | Degraded | Nobs species richness | Nobs seed density
Terrestrial | Tundra | Tundra | No | 43 | 40
Terrestrial | Tundra | Tundra | Yes | 3 | 3
Terrestrial | Forest | Boreal | No | 26 | 22
Terrestrial | Forest | Boreal | Yes | 11 | 11
Terrestrial | Forest | Temperate | No | 266 | 285
Terrestrial | Forest | Temperate | Yes | 160 | 144
Terrestrial | Forest | Tropical | No | 155 | 152
Terrestrial | Forest | Tropical | Yes | 136 | 135
Terrestrial | Grasslands & Savannas | Temperate & Boreal | No | 475 | 456
Terrestrial | Grasslands & Savannas | Temperate & Boreal | Yes | 340 | 265
Terrestrial | Grasslands & Savannas | Tropical | No | 51 | 45
Terrestrial | Grasslands & Savannas | Tropical | Yes | 33 | 31
Terrestrial | Mediterranean & Desert | Deserts & Xeric Shrublands | No | 107 | 83
Terrestrial | Mediterranean & Desert | Deserts & Xeric Shrublands | Yes | 33 | 35
Terrestrial | Mediterranean & Desert | Mediterranean Forests, Woodlands & Scrub | No | 129 | 121
Terrestrial | Mediterranean & Desert | Mediterranean Forests, Woodlands & Scrub | Yes | 69 | 61
Terrestrial | Arable | Mediterranean & Desert | NA | 48 | 43
Terrestrial | Arable | Temperate & Boreal | NA | 159 | 154
Terrestrial | Arable | Tropical | NA | 65 | 59
Transitional | Wetlands & Flooded Grasslands | Mediterranean & Desert | No | 40 | 23
Transitional | Wetlands & Flooded Grasslands | Mediterranean & Desert | Yes | 9 | 7
Transitional | Wetlands & Flooded Grasslands | Temperate & Boreal | No | 276 | 249
Transitional | Wetlands & Flooded Grasslands | Temperate & Boreal | Yes | 103 | 90
Transitional | Wetlands & Flooded Grasslands | Tropical | No | 50 | 42
Transitional | Wetlands & Flooded Grasslands | Tropical | Yes | 11 | 10
Aquatic | Aquatic | Aquatic | No | 51 | 44
Aquatic | Aquatic | Aquatic | Yes | 23 | 20
Total number of observations |  | 2872 | 2630

In this work, we use a hierarchical Bayesian modelling approach to investigate different components of the soil seed bank separately per biome, with ecoregion, habitat degradation (binary: undisturbed or degraded) and sampling effort (area sampled and number of sites) included as predictor variables (Table S2). Specifically, we investigate [1] species richness, finding highest richness in tropical forests and mediterranean ecosystems, but differences among ecosystems and biomes, reflecting our broad knowledge of global patterns in the aboveground vegetation; [2] seed density (m−2), finding highest values in wetlands and arable ecosystems; and [3] the impact of habitat degradation, finding consistently lower richness in degraded habitats, with patterns of density more variable across ecosystems. Importantly, we find that patterns of species richness in the soil seed bank diverged from patterns of density. That is, a high density of seeds in the soil seed bank does not necessarily translate to a large number of species, and this varies among ecosystems, biomes and in response to degradation.