RESUMO
BACKGROUND AND AIMS: Spatial variation in plant-pollinator interactions is a key driver of floral trait diversification. A so far overlooked qualitative aspect of this variation is the behavioural component on flowers that relates to the pollinator fit. We tested the hypothesis that variation in pollinator behaviour influences the geographical pattern of phenotypic selection across the distribution range of the oil-producing Krameria grandiflora (Krameriaceae). This variation mainly involves the presence or absence of flag petal grasping, which is only performed by representatives of Centris (Centridini, Apidae), an oil-collecting bee group highly associated with Krameriaceae pollination. METHODS: We quantified variation in floral traits and fitness and estimated pollinator-mediated selection in five populations at a large geographical scale comprising the entire species range. In each population, we sampled individual pollen arrival and germination as a fitness measure, indicating pollination success and pollination performance, which was then relativized and regressed on standardized flower-pollinator fit (flag-stigma distance), advertisement (sepal length) and reward (oil volume) traits. This generated mean-scaled selection gradients used to calculate geographical selection dispersion. KEY RESULTS: Unexpectedly, stronger selection was detected on the flower-pollinator fit trait in populations highly associated to the absence of the flag petal grasping. Geographical variation in selection was mainly attributed to differential selection on the flag-stigma distance generating a selection mosaic. This may involve influences of a spatial variation in pollinator behaviour as well as composition and morphology. CONCLUSIONS: Our results show the adaptive significance of the specialized "flag" petals of Krameria in the absence of the grasping behaviour and highlight the contribution of geographical variation in pollinator behaviour on flowers in driving selection mosaics, with implications for floral evolution, adaptation to pollinator fit and phenotypic diversity in specialized systems.
RESUMO
Encounters between flowers and invertebrates are key events for the functioning of tropical forests. Assessing the structure of networks composed of the interactions between those partners leads to a better understanding of ecosystem functioning and the effects of environmental factors on ecological processes. Gathering such data is, however, costly and time-consuming, especially in the highly diverse tropics. We aimed to provide a comprehensive repository of available flower-invertebrate interaction information for the Atlantic Forest, a South American tropical forest domain. Data were obtained from published works and "gray literature," such as theses and dissertations, as well as self-reports by co-authors. The data set has ~18,000 interaction records forming 482 networks, each containing between one and 1061 interaction links. Each network was sampled for about 200 h or less, with few exceptions. A total of 641 plant genera within 136 different families and 39 orders were reported, with the most abundant and rich families being Asteraceae, Fabaceae, and Rubiaceae. Invertebrates interacting with these plants were all arthropods from 10 orders, 129 families, and 581 genera, comprising 2419 morphotypes (including 988 named species). Hymenoptera was the most abundant and diverse order, with at least six times more records than the second-ranked order (Lepidoptera). The complete data set shows Hymenoptera interacting with all plant orders and also shows Diptera, Lepidoptera, Coleoptera, and Hemiptera to be important nodes. Among plants, Asterales and Fabales had the highest number of interactions. The best sampled environment was forest (~8000 records), followed by pastures and crops. Savanna, grasslands, and urban environments (among others) were also reported, indicating a wide range of approaches dedicated to collecting flower-invertebrate interaction data in the Atlantic Forest domain. Nevertheless, most reported data were from forest understory or lower strata, indicating a knowledge gap about flower-invertebrate interactions at the canopy. Also, access to remote regions remains a limitation, generating sampling bias across the geographical range of the Atlantic Forest. Future studies in these continuous and hard-to-access forested areas will yield important new information regarding the interactions between flowers and invertebrates in the Atlantic Forest. There are no copyright restrictions on the data set. Please cite this data paper if the data are used in publications and teaching events.