Natural disturbances, such as fires, droughts, or insect outbreaks, are key drivers of ecosystem dynamics and community structure1. Global change could exacerbate their severity and frequency worldwide with potential extensive impacts on biodiversity, ecosystems and human societies2,3. Understanding the effect of disturbances on the dynamics and structure of biodiversity is therefore more than ever a crucial issue in ecology. Yet, the high variability of local biodiversity trends in response to global changes asks for more integrative analyses, going beyond mere measures of species richness and accounting for the multiple components of the ecosystems4,5. Particularly, soil organisms are rarely included when synthesizing biodiversity trends in the face of disturbances, despite their recognized and well documented influence on multiple ecosystem functions (e.g. nutrient cycling) and nature contribution to people (e.g. carbon storage or depollution)6–9.
Most studies quantifying the effect of disturbances on biodiversity have focused on a single trophic or taxonomic group, often directly affected by the disturbance, like plants9. However, much less is known on how the effects propagate across trophic levels ultimately affecting the entire ecosystem. Plants and soil organisms are tightly linked through direct and indirect interactions, including mutualism, parasitism or predation, which promote the exchange and supply of nutrients and ensure multiple ecosystem processes6,7. Ignoring these trophic interactions and how resource deprivation in one trophic level can cascade to other levels may obscure the true consequences of disturbances for ecosystems10. Furthermore, misleading conclusions could be drawn if resulting disturbance effects are opposite between trophic levels11. Most natural disturbances cause immediate fluctuations in the quantity and quality of available soil resources1. Extreme winds can remove or deposit organic matter on the forest floor, while insect outbreaks increase soil nutrient inputs through defoliation and insect faeces and corpses. These local changes in basal resource availability can have rapid and important consequences on the abundance and diversity of primary producers (e.g. plants or nitrifying bacteria) and primary consumers (e.g. decomposers or herbivores), but also subsequently on the whole soil food web through bottom-up cascading effects12–14. However, there is a limited empirical knowledge on how these indirect effects propagate through the soil food web, and whether they vanish or amplify at increasing trophic levels12,15. In addition, changes in the abundance and diversity of organisms across the food web are likely to induce structural changes in the entire interaction network, potentially leading to alternative ecosystem states8,16,17. Thus, quantifying cascading effects of disturbances on ecosystems requires a holistic view of biodiversity with not only exhaustive sampling methods to capture all-in-end biodiversity, but also a suitable analytic approach to analyze change in trophic levels and interactions.
To meet this challenge, we combined the power of environmental DNA metabarcoding (eDNA)18 to obtain a nearly complete view of the belowground biodiversity, with a food web approach and network theory. Grouping species with the same trophic position (i.e. shared predators and preys/resources) in ecological networks facilitates the study of complex multitrophic communities. 19–21. In such an approach, the focus is not on species, but rather on trophic groups and trophic interactions. The definition of the trophic groups depends both on the resolution of the observation units (e.g. the taxonomic resolution) and the information available on their diet or trophic position22–24, and is also related to the ecological question. When looking at large-scale consequences of disturbance on biodiversity, a trade-off must be found between sufficient resolution to reliably and meaningfully measure cascading effects20,25, and resolution that is broad enough to avoid knowledge gaps and deal with heterogeneous taxonomic resolution in the data23,26. Once a food web is built, diversity can be measured within trophic groups (e.g. species diversity) and between trophic groups (e.g. trophic diversity or diversity of interactions), allowing the integration of ecological processes occurring at different dimensions of the food web (e.g. competition and predation)27,28. For this, network theory provides appropriate metrics to describe and compare the diversity and structure of ecological networks, accounting for both group abundances and interactions29,30.
Here, we study the effect of moth outbreaks on soil food webs of subarctic birch forests in Northern Fennoscandia. These forests have experienced moth outbreaks of unprecedented scale and severity in recent decades, which have led to a sudden vegetation change from birch forests with understory dominated by dwarf shrubs to grass-dominated systems associated with high tree mortality31–34 (Fig. 1). Moth outbreaks is a good model for assessing the cascading effects of disturbance on soil food webs, as the larvae only attack the foliage of the dominant primary producers, i.e. the birch tree (Betula pubescens), and some abundant species of erect and dwarf shrubs in the understory layer (e.g. Betula nana, Empetrum nigrum, Vaccinium spp.). In parallel, soil organic matter is enriched through dead plants and N addition from larval faeces and corpses35,36. We can therefore assume that impacts on the whole soil food web arise from bottom-up effects from changes in the vegetation and basal resources to the other trophic compartments 12. Drastic shifts in the composition of biological communities following defoliation have been already reported in these nutrient-limited soils where the dominance of the allelopathic dwarf shrub Empetrum nigrum in the understory leads to regressive succession that may inhibit soil microbial activity, organic matter decomposition, and thus nutrient availability37–39. These shifts correspond to a replacement of Empetrum nigrum by a grass Avenella flexuosa32 with subsequent effects on the diversity and abundance of organisms directly relying on plants, including vertebrate herbivores31, birds40, saproxylic beetles36, and fungal communities41,42. However, we still ignore whether moth outbreaks induced indirect effects across the soil food web, whether these effects are of comparable magnitude to those observed for vegetation, and finally, whether these effects have significant consequences on trophic interactions and ultimately on the whole soil food web structure.
We used eDNA data obtained from 86 soil samples from two well-studied areas in northeastern Norway (i.e. Tana and Kirkenes). This study design allowed for appropriate pairwise comparisons between coupled undamaged and defoliated forest plots based on well-documented defoliation patterns from both remote sensing and field methods (Fig. 1). The sampling design aimed at capturing the environmental heterogeneity at different spatial scales of the landscape within these areas (see methods). Soil organisms distribution is influenced by different factors depending on the spatial scale of observation (from cm to km) and accounting for such a scaling is thus essential43,44. Different spatial scales were included in the statistical models used for analyses. By mobilizing currently available ontologies and knowledge about soil trophic groups, we classified the retrieved soil organisms, from microorganisms to macroinvertebrates, into 9 broad feeding categories that we call trophic classes. These classes were further subdivided in 37 trophic groups by separating distant phylogenetic groups that might have a different set of prey/predators (e.g. bacterivore mites and bacterivore nematodes), or groups that differ in their resource acquisition strategy (e.g. different types of mycorrhizal and saprotroph fungi; Supplementary Table 1). Next, we added interactions between trophic classes (low resolution, Fig. 2a) and between trophic groups (high resolution, Fig. 2b, Supplementary Fig. 1) according to their associated feeding preferences and additional empirical knowledge from scientific literature and interaction databases. When interaction knowledge was lacking, we generalized information taxonomically at different levels depending on the organisms (see methods). The metaweb45 obtained at two levels of resolution, corresponding to the potential knowledge-based soil food web of the study area, was then used to characterize the local soil food webs based on the trophic classes or groups detected locally in each soil sample. Nodes of the local food webs corresponded to the local relative abundances of the groups or classes. In eDNA metabarcoding studies, changes in the abundance/biomass of an individual taxon may be inferred, in some extents, from changes in their relative abundances across samples, although this correspondence can be noised by different biological or technical factors (reviewed in 18). However, some taxon can exhibit higher gene copies than others, making these changes in relative abundance more difficult to compare across groups contrary to other abundance standardized measures such as biomass. We used a measure of relative abundance varying from 0 (when the group was absent) to 1 (when the group was at its maximum observed abundance), in order to have comparable values between the different groups. The trophic class resolution corresponds to what is commonly used in soil food web ecology (e.g.20,25), but we additionally included the trophic group resolution because a finer resolution is needed to capture specific effects of disturbance on groups that are hidden at a coarser resolution. For instance, different types of mycorrhizal fungi like arbuscular mycorrhizal fungi and ectomycorrhizal fungi may have opposite responses to tree defoliation, the former increasing and the later decreasing in their proportion following disturbances46.
Using this holistic approach, we tested several hypotheses about the cascading effects of moth defoliation on the local soil food webs. First, (H1) moth defoliation changes the diversity in MOTUs (Molecular Operational Taxonomic Unit) and the relative abundances of most trophic groups. We expected positive effects on most decomposers and their consumers through the impulse in soil resources availability47,48 from both moth outbreaks and the decreased abundance of the allelopathic species Empetrum nigrum. In parallel, we expected negative effects on e.g. ectomycorrhizal and ericoid mycorrhizal fungi, as the result of the decline of birch and ericaceous shrub roots. Second, (H2) the effect of defoliation is strongest for primary consumers and decomposers who are directly affected by changes in basal resources availability and plant composition, and then decreases toward higher trophic levels (attenuation of the effects). Third, (H3), moth defoliation changes the overall structure of the local soil food webs10,49.
To test H1 and H2, we used a multilevel linear model for each trophic group using defoliation as a predictor and accounting for the spatial hierarchical structure of the sampling design (see methods). To test H3, we used network diversity indices29 to study changes in the structure of local food webs. First, we analysed whether defoliation caused a change in the structure of soil food webs using a measure of dissimilarity of node and link compositions29 at fine and coarse resolutions (trophic groups and classes respectively). Next, we compared the local diversity (α-diversity) of the soil food webs between undamaged and defoliated forests using the exponential of Shannon entropy applied to nodes and links abundance, at different resolutions.