Abundance-driven processes best predict plant-butterfly network structure across a forest-to-agriculture gradient

Land-use change threatens pollination services globally, yet interaction networks may persist through ecological processes that buffer habitat modification. We studied plant-butterfly nectar-feeding interaction networks across a forest-to-agriculture gradient in temperate Mexican forests, comparing juniper-oak forests, eucalyptus forests, and croplands over one year. We evaluated network structure and used probabilistic models to assess the relative predictive power of neutral processes (species abundance), phenological overlap, and morphological matching in explaining observed interaction patterns. We recorded 3,229 interactions among 76 species (34 plants, 42 butterflies) with moderate specialization. Networks showed significant nestedness and modularity across habitats, though weighted nestedness was lower than expected. Model selection revealed that species abundance primarily predicted network structure, with phenology becoming important in croplands where floral resources are temporally concentrated. Generalist butterflies, particularly Battus philenor philenor , maintained consistent structural roles across habitats, demonstrating functional stability. Pieridae and Papilionidae contributed positively to nestedness through generalist interactions, while Hesperiidae and Lycaenidae reduced it through specialized associations. Our findings suggest that abundance-driven processes best predict nectar-feeding interaction patterns in modified landscapes, though we emphasize that structural persistence does not guarantee pollination effectiveness. Conservation strategies could prioritize abundant native flowering plants and habitat heterogeneity that may support generalist populations, while recognizing that network topology may not fully capture functional aspects of pollination such as pollen transfer efficiency and plant reproductive success.