What co‐occurrence networks can and cannot tell us about soil ecology

Abstract

Co-occurrence network analyses are widely used to explore soil microbial communities, providing a framework to describe associations among soil taxa and their potential responses to environmental gradients or treatments. These networks are ofteninterpreted as representations of ecological interactions yet, the ecological (and statistical) robustness of such inferences dependson whether the network in question truly reflects underlying biological associations rather than just spurious correlations oreven methodological artifacts. In this opinion paper, we examine the premise, strengths, and limitations of co-occurrence network analyses and discuss methodological challenges associated with data compositionality, sample size, taxonomic resolution,and the spatial and temporal scales at which interactions occur in soil. We highlight the importance of robust null models forstatistical validation of network patterns to identify genuine structures and their ecological implications. We finally synthesizeapproaches for integration between network science and soil ecology to enable co- occurrence networks move from descriptive analyses to understanding the mechanisms that structure soil biodiversity and ecosystem functioning.