Introduction
Soil biodiversity encompasses a complex network of interactions among functionally and trophically diverse organisms, playing a vital role in supporting ecosystem functions and services such as carbon sequestration, organic matter decomposition, and enhancing plant performance and resistance to pests and stress (Bardgett & van der Putten 2014; Delgado-Baquerizo et al. 2020). Over the past decades, our understanding of the spatial distribution of soil taxa has improved, particularly for specific groups like bacteria (Fierer & Jackson 2006; Delgado-Baquerizo et al. 2018) and fungi (Tedersooet al. 2014; Davison et al. 2015). However, our current knowledge primarily relies on a few specific taxa, while the broader ecological processes shaping the whole soil communities and their organisation into interaction networks remain poorly understood. This knowledge gap stem from challenges associated with capturing the intricate spatial heterogeneity inherent to the soil environment (Ettema & Wardle 2002), and the methodological complexities of sampling and analysing the vast diversity of the soil biota (Geisen et al.2019). These limitations have hindered the application of existing theoretical frameworks on soil biota (Thakur et al. 2020). Advancing research in this direction promises to unravel the ecological processes that structure soil biodiversity and to predict the impacts of global change on terrestrial ecosystems (Soliveres et al. 2016; Eisenhauer et al. 2021).
Gaining insights into the organization of soil communities necessitates studying spatial, environmental, and biotic drivers and this poses challenges (Decaëns 2010; Münkemüller et al. 2020; Eisenhaueret al. 2022). First, capturing the full complexity of soil systems requires accounting for their inherent spatial heterogeneity from landscape to local scales (Ettema & Wardle 2002; Thakur et al. 2020). At the landscape scale, dominant habitat types, such as grasslands or forests, are crucial for soil taxonomic and functional composition (Fiore-Donno et al. 2020; Arribas et al. 2021; Sepp et al. 2021). Consequently, the heterogeneity between habitats within a landscape can give rise to mosaic patterns of soil communities. Moreover, the spatial connectivity of habitats and the varying dispersal capacities of soil organisms can influence the local composition and structure of soil communities (Arribas et al.2021). But even at the local scale of a few meters, variable abiotic conditions, including microclimate and soil physico-chemical properties, along with stochastic processes, can lead to pronounced differences in soil communities (Ramirez et al. 2014; O’Brien et al.2016; Zinger et al. 2019). The structure of the vegetation within an habitat or its taxonomy or functional composition can also affect the abundance and diversity of different soil taxa or trophic groups (Noguerales et al. 2021; Calderón‐Sanou et al. 2022; Ganault et al. 2022). Effectively disentangling the effects of spatial, environmental and biotic processes necessitates a sampling design that encompasses multiple spatial scales, ranging from the heterogeneity between habitats to small-scale soil variations.
Second, comprehensive sampling and cross-taxa analysis are required to capture the high diversity in soils. Recent advancements in monitoring techniques, such as environmental DNA metabarcoding (eDNA)(Taberletet al. 2012), offer cost-effective means to obtain extensive data on soil biota at large spatial extent (Taberlet et al. 2018; Geisen et al. 2019). To more concisely present the different taxa they can be classified into trophic groups that share similar resources or prey (Eltonian niche; Elton 1927), e.g., fungivorous nematodes or photoautotrophic bacteria (Louca et al. 2016; Potapov et al. 2022). To better comprehend the complex interactions, we can construct food webs that consider multiple trophic groups (nodes) and their linkages across trophic levels (links) (Thompson et al.2012; Gravel et al. 2019). However, constructing food webs from large eDNA datasets is not without challenges. It entails handling and classifying thousands of sequences into trophic groups, and subsequently linking these groups based on known interactions. Obstacles include the resolution limitations of DNA markers for species-level assignment, the incompleteness of sequence reference databases, inconsistent terminology in trophic ecology across soil taxa (Hedde et al. 2022), and the sparse trophic information available for soil organisms in literature and repositories (but see Potapov et al. 2022). Nevertheless, the application of artificial intelligence tools now offers the potential for automating classification tasks (Compson et al. 2018; Le Guillarme & Thuiller 2023).
Third, if we are to understand soil biodiversity, not as a group of independent taxa, but rather as food webs, we need the appropriate methods and frameworks to investigate how ecological processes act on soil trophic groups and their interactions. The mono-trophic community assembly framework (Keddy 1992; Thuiller et al. 2013) can be extended to food webs using the metaweb concept (Dunne 2006). At the regional scale, the metaweb represents the potential food web encompassing all species (or trophic groups) from the regional pool and their potential interactions (Fig.1c). At the local scale, realised local food webs emerge as subsets of the metaweb due to ecological filtering of species (Bauer et al. 2022). This filtering process can be influenced by spatial filters like landscape barriers, limiting dispersal (Peay et al. 2010), abiotic factors filtering out species lacking physiological adaptations (Maaß et al. 2015; Glassman et al. 2017), and biotic interactions such as mutualisms, herbivory and predation (Vályi et al. 2016). By using the metaweb concept, we can evaluate the effects of ecological filters not only on the taxonomic composition of the local community but also on its network structure (i.e., composition of interactions). Important biotic interactions in the soil are via plant-soil feedbacks, where the taxonomic or functional composition of plant communities influences soil food webs and vice-versa (Kardol & De Long 2018; Kardol et al.2018). Moreover, trophic interactions within the food web can elucidate its local structure and composition (Thakur & Geisen 2019). We anticipate that interactions between taxa or groups within the food web should result in co-variation across environmental and spatial gradients. For instance, environments characterized by decomposer communities dominated by bacteria (rather than fungi) should promote the dominance of bacteria consumers and related higher-level consumers within the food web (Moore & de Ruiter 1991), see e.g. Martinez-Almoynaet al. 2022. Null models can be used to test the significance of such co-variation in the assembly of ecological communities (Carusoet al. 2022).
Here, we analysed the spatial variation in soil food web structure, encompassing trophic group and interaction composition. We used data from all over the French Alps including lowland forests and high-altitude grassland ecosystems. To represent their spatial heterogeneity, we employed a stratified and nested sampling design, consisting of 24 elevational gradients with multiple plots at varying altitudes (Fig. 1a). We measured soil biodiversity in 418 soil samples, using environmental DNA (eDNA). Using an ontology-based data integration pipeline, combining multiple databases with existing knowledge on the trophic habits of soil organisms (Le Guillarme et al. 2023; Le Guillarme & Thuiller 2023), we constructed a metaweb comprising 55 soil trophic groups with 383 potential trophic interactions (Fig.1b). Finally, we applied the metaweb framework to characterise the importance of ecological filters on soil food webs (Fig.1c), using a combination of network diversity indices based on Hill numbers (Ohlmann et al.2019) and associated null models. We quantified local variations in soil food web structure both between and within habitats, addressing three specific questions: (Q1) Do different habitats differ in soil food webs?, (Q2) Do trophic interactions contribute to variations in group abundances between habitats?, and, (Q3) How do abiotic conditions, spatial factors, and plant communities shape soil food webs within habitats?