From eDNA to food webs: AI ‐enabled metaweb construction for lakes which captures non‐random species assemblies
Year: 2026
Type: article
Abstract: Food-web structure reveals ecological processes that community composition alone cannot. The rapid uptake of environmental DNA (eDNA) sampling has greatly expanded freshwater biodiversity data, creating a need for equally scalable methods to assess trophic structure, particularly in under-represented regions and among overlooked small-bodied consumers. Metawebs-regional inventories of trophic interactions used to infer local realised food webs-offer a promising solution. We developed an AI-assisted workflow to infer predator-prey interactions and used it to build a regional metaweb for New Zealand lakes from eDNA data. From a regional pool of 312 freshwater taxa, we produced a high-confidence-link (HCL) network and a broader all-consumers (AC) network. We validated the workflow against expert assessment and compared local networks from 15 lakes with null models to test for non-random structure. Observed communities had significantly higher connectance and mean trophic level than random assemblages, indicating that trophic interactions constrain community assembly. These patterns were stronger in HCL than AC networks. Food-web metrics-especially modularity-were also more sensitive to environmental gradients than species richness, providing insight beyond conventional biodiversity measures. This scalable, cost-effective framework infers trophic structure from eDNA data and can support ecosystem monitoring, management and prioritisation of lakes for targeted investigation.
Source: Journal of Animal Ecology
Language: English
Subfield: Ecology
Field: Environmental Science
Domain: Physical Sciences
SDG: Life on land
Open Access status: hybrid
Estimated APC paid: $3,610
OpenAlex ID: W7236849966
