How different link weights affect the structure of quantitative flower-visitation networks

Abstract

Despite the considerable work carried out to assess the structure of weighted and unweighted mutualistic networks, little is known about how different ways to measure the weight of interactions can influence network parameters used to describe such structures. The use of an appropriate 'link weight' is especially important if we want to move toward a more functional perspective in ecological network studies. Here, we evaluated how the use of five different link weights, starting with the simplest one - the number of visits - and including progressively information on total census time, number of flowers observed, number of flowers visited and total flower abundance of each species in a plant community, influences widely used descriptors of network structure. We analyzed different network-level properties: weighted nestedness, flower-visitor generality, plant generality, complementary specialization H 2', interaction evenness and interaction diversity, as well as species-level parameters: specialization index d' and strength. We built two quantitative flower-visitation networks from two different communities sampled during two consecutive years, in which we also measured independently flower abundance of each plant species. Results showed that the type of link weight used can significantly alter network structure. A clear trend toward more specialized interactions (decreased flower-visitor generality, increased H 2' and d'), higher heterogeneity in the frequency of interactions (lower evenness and diversity of interactions) and higher weighted nestedness was found as interaction weight gained in complexity. However, standardizing only for total censusing time had only weak effects on network parameters. Our findings highlight the importance of carefully considering the most appropriate link weight for each ecological network study, emphasizing that comparisons across networks that use different weights might lead to flawed results and thus to ecological misinterpretations of network structures. © 2012 Gesellschaft far kologie.