Effect of temporal data aggregation on the perceived structure of a quantitative plant-floral visitor network

Abstract

Seasonal turnover in plant and floral visitor communities changes the structure of the network of interactions they are involved in. Despite the dynamic nature of plant–visitor networks, a usual procedure is to pool year-round interaction data into a single network which may result in a biased depiction of the real structure of the interaction network. The annual temporal dynamics and the effect of merging monthly data have previously been described for qualitative data (i.e. describing the occurrence of interactions) alone, while its quantitative aspect (i.e. the actual frequency with which interactions occur) remain little explored. For this, we built a set of 12 monthly networks describing year-round plant–floral visitor interactions in a 30-hectare planted forest and its adjacent agricultural landscape at Bahauddin Zakariya University Multan, Pakistan. A total of 80 plant and 162 insect species, which engaged in 1573 unique interactions, were recorded. Most network properties (particularly the number of plants, visitors and unique interactions) varied markedly during the year. Data aggregation showed that while animal species, plant species, unique interaction, weighted nestedness, interaction diversity and robustness increased, connectance and specialization decreased. The only metric which seemed relatively unaffected by data pooling was interaction evenness. In general, quantitative metrics were relatively less affected by temporal data aggregation than qualitative ones. Avoiding data aggregation not only gives a more realistic depiction of the dynamic nature of plant–visitor community networks, but also avoids biasing network metrics and, consequently, their expected response to disturbances such as the loss of species.