Forest cover and landscape heterogeneity shape ant–plant co-occurrence networks in human-dominated tropical rainforests

Context: The effects of habitat loss and fragmentation on biodiversity involve a series of mechanisms and processes that cannot be studied in isolation, mainly because human-modified landscapes are spatially heterogeneous. Despite the recent evidence on how habitat quality and landscape structure may regulate species composition and community structure, there is no empirical information on how ant–plant co-occurrence networks respond to landscape changes. Objectives: In this study, we used tools derived from landscape ecology and graph theory to model how habitat loss and forest fragmentation affect ant–plant co-occurrence networks at two different spatial levels (local and landscape) in a human-modified tropical rainforest in Mexico. Methods We sampled ant–plant co-occurrence networks in 16 landscapes and calculated species and co-occurrence diversity as well as the specialisation of the networks. For each landscape, we measured a series of biotic and abiotic variables at the local level and another set of variables at the landscape level. Results: We found that the landscape context (i.e. forest cover and landscape heterogeneity) was a better predictor of plant diversity and diversity of ant–plant co-occurrences compared with local characteristics. Moreover, we also observed that network specialisation was positively related to the amount of forest cover in landscapes. Conclusions: In short, our results highlight the importance of studying the diversity of species and co-occurrence networks within a landscape approach to predict the impact of habitat loss and fragmentation at different spatial levels and contexts.