Network architecture and vulnerability of macroalgal epiphyte-host interactions in Caribbean coastal ecosystems: implications for marine conservation

Background and Aims Tropical coastal marine ecosystems depend critically on macroalgal epiphyte–host interaction networks to sustain biodiversity, primary production, and ecosystem services, yet the architecture and vulnerability of these networks remain poorly characterised in the Caribbean. This study aimed to characterise the topological structure of the epiphyte–host interaction network in Cuba's coastal marine ecosystems, model the ecological determinants of epiphytic richness, quantify niche overlap among hosts, assess structural robustness to host species loss, and develop an Integrated Community Vulnerability Index (Vtotal) to guide conservation prioritisation. Methods A bipartite network was constructed from a comprehensive synthesis database comprising 62 host taxa and 321 epiphytic macroalgal species across eight coastal habitat types. Network topology was analysed using degree distribution modelling, Louvain community detection, and centrality metrics. Host habitat distributions were inferred via a hierarchical Bayesian model validated against 18 hosts of known distribution. Epiphytic richness was modelled using a Bayesian negative binomial regression with inferred habitat breadth and betweenness centrality as predictors. Niche overlap was assessed with Jaccard similarity and Monte Carlo permutation tests. Structural robustness was evaluated through sequential host-removal simulations under three strategies. Vtotal was calculated by integrating habitat environmental vulnerability with network-structural irreplaceability. Key Results The network comprised 396 nodes and 1,089 edges (connectance = 0.0315), with degree distribution best described by a truncated power law. Eighteen modules were identified (Q = 0.424). Betweenness centrality (β = 0.957) was a stronger predictor of epiphytic richness than habitat breadth (β = 0.524). Overall Jaccard similarity was extremely low (mean = 0.027), indicating high functional complementarity. Targeted host removal caused network fragmentation after only 14.6% of hosts were removed, versus 64.6% under random removal. Rhizophora mangle, Thalassia testudinum, and Digenea simplex emerged as irreplaceable structural hubs. Mangrove module M9 attained the highest Vtotal (4.264). Conclusions The epiphyte–host network is architecturally efficient yet ecologically fragile, with a disproportionate dependence on a small set of engineering host species whose loss would trigger non-linear cascades of secondary epiphyte extinctions and ecosystem service degradation. The Integrated Vulnerability Index provides a transferable, interaction-based framework for conservation prioritisation in tropical marine systems facing accelerated environmental change.